As outlined in the Research Methodology section, Principal Component Analysis (PCA) performs dimensionality contraction of data and hence is a valuable tool for determining whether scale items within a metric can be reduced into a set of factors, each effectively becoming a subscale in its own right. The main purpose of the analysis is to explain as much of the variance in the variables as possible using as few components (factors) as possible. However the process also enables an examination of the resulting subscales to be conducted which can be introduced into the analysis of the data to help in understanding how these components of the main metrics might interact with each other. This is important because by exploring these interactions, the ways in which aspects of dyslexianess might affect aspects of academic confidence may be more clearly revealed and hence, better understood.
Hence both the Dyslexia Index Profiler and the Academic Behavioural Confidence Scale have been reduced through PCA into a set of factors (components) and these have been assigned namelabels that reflect which dimensions of their parent metrics are their respective contributors. The software application SPSS has been used to work through the calculations and the report that follows distils the outputs generated from running these processes. As reported in the previous section, Sander & Sanders also applied a process of factorial analysis to their original, 24item Academic Behavioural Confidence Scale and later, to their revised, 17item Scale. With the exception of one study found to date (Corkery, 2011), all others that have used the ABC Scale in their research have utilized the Sander & Sanders factor structure for their analysis, had they chosen to explore their findings in the greater detail that the use of ABC factors permits. The original, together with Corkery's ABC factor structures have been investigated as part of my analysis development and reported elsewhere in this thesis (here), the outcome of which suggested that there would be merit in developing a factor structure that is unique to this project because the process may reveal an alternative set of projectspecific factors for the Academic Behavioural Confidence Scale which could be more relevant for exploring the interrelationships between components of academic confidence and components of dyslexianess that are the focus of this enquiry.
In order to run a principal component analysis there are a number basic assumptions which need to be fulfilled:
The first is met through the study design for this project which is that the variables that comprise the two comparator metrics are continuous. This has been incorporated into the questionnaire design through use of continuous range input sliders as the mechanism for acquiring research participant response scores for the two scales, Dyslexia Index and Academic Behavioural Confidence which have been developed specifically to enable parametric statistical processes to be used in the data analysis. As alternatives to the more conventional, fixed anchorpoint Likertstyle scales which would have produced discrete data values leading to data outputs being arbitrarily coded, this design feature of the data collection tool thus supports a more robust statistical analysis.
The second assumption is that there needs to be evidence of linear relationships between the variables that comprise the scales. This is because PCA is based on Pearson correlation coefficients and it is considered that a scaleitem variable that presents a correlation of r ≥ 0.3 with at least one other scaleitem variable is worthy of inclusion in the analysis (Hinton et al, 2004). In the case of the Dyslexia Index, there are 20 scale item variables and for Academic Behavioural Confidence there are 24. An analysis of the intervariable correlation matrix for both metrics showed that for Dyslexia Index, of the 190 possible correlation outcomes, 80 returned a correlation coefficient of r ≥ 0.3 with all variables bar one returning at least one correlation of r ≥ 0.3 with any other variable. For the Academic Behavioural Confidence Scale, of the 300 possible correlation outcomes, 138 returned a correlation coefficient of r ≥ 0.3 with all variables returning at least one correlation of r ≥ 0.3. Hence it is considered that the second assumption for applying a PCA to both metrics is met.
The fourth assumption tests the null hypothesis that there are no correlations between any of the variables. If this null hypothesis were accepted, this would be indicating that the set of variables may not be reducible to a smaller number of components (factors), which is after all, the rationale for applying a principal component analysis to the data. The measure for testing this null hypothesis is Bartlett's Test of Sphericity and the output being sought is a significant result (p < 0.05) which will enable the null hypothesis that there are no correlations between any of the variable to be rejected. When applied to the Dyslexia Index metric, the test returned an output of p < 0.005, and when applied to the ABC Scale, the output was also p < 0.005. Hence for both metrics, the null hypothesis that there is no correlations between the metrics' variables is rejected, which means that there are correlations between the variables and therefore justification for running the PCA is present
Hence the preliminary assumptions have been met for running a principal component analysis on both the Dyslexia Index and on the Academic Behavioural Confidence Scale for the data collected in this enquiry.
The key outcome of the factorial analysis of both metrics is that it has enabled a Dx Factor X ABC Factor matrix to be contsructed which, as expected, has revealed highly interesting and meaningful relationships between the components of the two metrics and this has contributed significantly to the discussion and final conclusions. The matrix and a more detailed report is presented below.
Recall that the Likert scale that is attempting to evaluate levels of dyslexianess, the Dyslexia Index (Dx) Profiler, is a 20item scale comprising scale items which attempt to evaluate the extent of each respondent's acquiescence with statements concerning specific aspects, that is, dimensions, of study, academic learning management and to an extent, learning history. The scale items were assembled in the questionnaire in a random order (left) and respondents were required to adjust the position of a slider control along a scale from 0 to 100% to register their degree of acquiescence with each of the statements. Each statement had its own slider. This process for recording participants' responses has been used throughout the earlier sections of the questionnaire. The design rationales are described in the previous section, 'Research Design' but the principal aim has been to mitigate the conventional Likertstyle scale limitations imposed through fixed anchorpoint selectors by generating a data output that can be considered as a measure of a continuous variable.
It is important to emphasize that the main function of factorial analysis in this project is to enable the factor subscales that emerge out of the Dyslexia Index metric to be related factorbyfactor to data acquired through the other main metric, Academic Behavioural Confidence (ABC) and its own factors (note that 'component' and 'factor' are used interchangeably throughout). The process of PCA determines that through interpretation of the output, scaleitem dimensions of the Dyslexia Index scale will be assigned to only one factor  hence none will have crossfactoral influence, and likewise for ABC. Note that there is an element of 'best option' choice involved in this process as usually dimensions load on to more than one factor and a decision has to be made about determining which balance of options is the most interpretable. More of this below.
A significant further element of PCA is that it enables statements to be constructed which explain the proportion of variance that can be explained by each component and this can be usefully interpreted as the strength of impact that a component has on the complete results. This has been discussed in more depth in the previous section 'Research Design'. [Note to me: expand this a little in the RD section]
PCA has also been used to help to identify scale items that might be considered as redundant  that is, are not contributing to the evaluation of the construct in a helpful way and hence might be discarded. A process of inspection for scaleitem redundancy has already been conducted through interpretation of the Cronbach's Alpha (α) internal reliability consistency analysis which has been thoroughly reported in the previous section, Research Design, however it is useful to provide a recap here as the outcomes contribute to this report on how the factors for Dyslexia Index (and ABC) have been established.
For the 20item Dx scale used to collect participant data in the main research questionnaire, Cronbach's α = 0.842 which appears to be indicating a strong level of internal reliability consistency. As stated earlier, according to Kline (1986) however, a value α > 0.7 of does not necessarily strengthen this further and may instead be suggesting that some scale items can be removed to have little impact on the αvalue and hence maintain the level of reliability of the scale without sacrifices possibly perceived as resulting from the subsequent reduced data acquisition.
When the potentially redundant Dx scale items were identified through this analysis and removed, the resulting 16item scale returned a value of α = 0.889 which is indeed indicating that the internal consistency reliability of the scale is maintained and hence suggests that discarding these 4 scale items from the Dyslexia Index Profiler, thus reducing it to a 16item scale, may have little impact on the overall Dx values.
The first iteration of this process which has regenerated 16point Dyslexia Index Scale warrants further testing and this may be the focus of a followup study in due course, one element of which will be to explore the impact on α when combinations of the possibly redundant items are withdrawn. Suffice to say that effect size results for differences in ABC between research subgroups established through the 20point Dx scale compared to those established through the reduced, 16point Dx scale were only marginally different and so for this primary analysis, the full, 20point scale has been used, pending the outcomes of the suggested followup study later.
Principal Component Analysis will produce as many components (factors) as variables and the process will 'explain' all the variance in each of the variables if all of the components are retained. Clearly the objective of the process is to decide how many factors are worth retaining in the final solution so that as much of the total variance as possible can be explained through all of these factors. Using the Eigenvalue1 extraction criteria (Kaiser, 1960) which is the default setting in SPSS on account of this criteria being claimed as the most popular, five components (factors) emerged from the analysis output in SPSS which between them accounted for 60.4% of the total variance. Respectively these factors accounted for 31.7%, 9.9%, 7.6%. 6.0% and 5.3% of the total variance for Dyslexia Index. Inspection of the scree plot that was generated as part of the output suggested that retaining these 5 components (factors) would be appropriate although it can be seen both from the scree plot and from the table (right) that it is possible that a sixfactor solution may be equally applicable because the initial eigenvalues for components 5 and 6 were both very close to 1, (1.06, 0.988), or even a fourfactor solution may be the most appropriate as the eigenvalue for the fourth component in the 5factor solution stood at a value of 1.20. To explore this, the principal component analysis was run twice more, firstly with a forced extraction of six components and secondly with four components. The sixcomponent extraction produced an output where the last two components comprised just one dimension each and so this analysis was dismissed. The fourcomponent extraction produced a highly similar output to the original, fivefactor solution with just two dimensions being alternatively placed into different factors. Hence it was concluded that the fivefactor solution can be accepted as a reasonable factor structure for the metric Dyslexia Index.
In interpreting factor analysis outputs, the Table of Communalities is the first result that is useful and from this, the groupings of dimensions into factors makes its first appearance. Five factors emerged and these have been labelled in keeping with the flavour of the dimensions which comprise them respectively:
The communalities are the proportions of each of the variables' variances that is accounted for by the principal component analysis. So for example, for the first dimension in the table below 3.20: I get really anxious if I'm asked to read 'out loud', the communalities extraction value of 0.573 indicates that 57.3% of this dimension's variance can be explained by the factors (components). The loading is the correlation between the variable and the factor and this is(are) the figure(s) presented in line with each dimension in the respective factor column. According to research convention, serious attention is paid to loading factors of > 0.32 (Dewbury, 2004, p309) with Dewbury citing a reference to an earlier work by Comry & Lee (1992) which proposes that a loading of > 0.71 is 'excellent'. Note that in the table below, although loadings are calculated for all dimensions in all factors, only factor loadings > 0.3 are presented to enable the table to be more easily assimilated. Hence the row of data for dimension 3.20 only shows the one value of 0.829 for a loading onto Factor 1, Reading, Writing, Spelling because the loadings onto the other four factors are less than 0.3, and so forth for all dimensions and factors. The communalities extraction figure of 0.573 is thus the proportion of this dimension's variance that can be accounted for by all of the factors.
These communalities are reported alongside the Rotated Component Matrix which is a table that groups the 20 dimensions into the 5 components/factors, where in each component, dimensions are listed in descending order according to the loading onto each variable ( = dimension). The table indicates 'rotated' components which is the mathematical process that places the factors in the best (geometrical) position to enable easier interpretation. SPSS uses a Varimax' rotation, this being an 'orthoganal' process meaning that factors are forced to be independent of each other (rather than taking into account correlations between them).
What emerges from this matrix (below) is that the factor structure is not quite as simple as would have been desirable because some dimensions load onto more than one factor, given the convention of a loading > 0.3 is indicating an influence that should be taken seriously. However, where this occurs, the troublesome dimension has been assigned to the factor onto which its loading is greatest  that is, where there is the greatest correlation between the dimension and the factor. Kline (1986), amongst other consulted, suggests that more often than not a single, simple factor structure is elusive and it should remains the task of the researcher to establish the most appropriate interpretation of the analysis that makes sense in the context of the project. This factor analysis for the Dyslexia Index Profiler seems reasonable and so for the purposes of the remaining analysis of the data collected in this project, this factor structure with the 20point Dyslexia Index comprising 5 factors will be retained.
Dyslexia Index (Dx) Factors by Research Subgroup
The factorial analysis for Dyslexia Index has enabled strongly visual radar charts to be constructed which present an overview of the distribution of Dx factor values for students in each of the three, key research subgroups in the study. The three charts (below) display the factor profile for every student in these subgroups with profiles overlaid on to each other so that composite profile maps can be constructed. In ways that are much easier to spot than through inspection of the full data tables (below) from which the profile maps have been derived, these graphical representations of the five factor values for both each, and all students in each subgroup reveal stark differences between the factor profiles of nondyslexic students to those who had declared their dyslexia and those apparently nondyslexic students who, through the sifting process that the Dyslexia Index Profiler has enabled, are presenting levels of dyslexianess of comparable levels to the students with dyslexia. It is apparent that the collective profile maps for students in the TEST research subgroup, RG:DNI are highly similar to the collective profile maps for students in the CONTROL subgroup, RG:DI600 where students in both of these subgroups present a Dyslexia Index value of Dx > 592.5 which is indicating that highly similar combinations of factor values are present in both subgroups, implying stron dyslexianess similarities between students with known dyslexia and, as is becoming increasingly apparent, students with potentially unknown dyslexia. Both of these are significantly different from the collective profile maps for students in the BASE subgroup, RG:ND400, (Dx < 400). It is clear to see the noticeable skew in the chart for the strongly nondyslexic students away from the two Dyslexia Index factors, 'Reading, Writing, Spelling' and 'Thinking, Processing' but also more generally this profile maps indicates reduced Dyslexia Index Factor values overall for students in this group in comparison with students with declared dyslexia and those nondyslexics who appear to be presenting dyslexic characteristics. Aside from being highly revealing of differences at the factorial level which will be discussed later, it is argued that this representation further demonstrate the validity of the Dyslexia Index Profiler as an effective discriminator for the purposes of this study. Each of these radar charts can be viewed in deeper detail in conjunction with each of the full data tables below.
The data tables collected here show a detailed summary of the factor values for every student in each of the three research subgroups of main interest. Two visualizations of each set of data values are available but it is clear to see how much easier it is to absorb the results collectively when presented as a radar chart in comparison to the scatter diagram of the same data which was the first attempt at visualizing the data.
[Respondents highlighted in green text in the tables are linked to their additional written commentaries submitted in the questionnaires and themes that have emerged in these will be included in the discussion of the results in the next section.]
The matrix below shows a condensed view of the complete set of (same Dx Factor mean value) x (same Dx Factor mean value) combinations for each pair of research subgroups. The purpose of this summary table is to present evidence of the significant or not significant differences between the Dx Factor means across the research subgroups as a first step towards a more detailed level of scrutiny of the subgroup differences in dyslexianess that have emerged.
Firstly, it is important to note that for all Dx factors, the ttest reports no significant differences between the samples means for each factor when comparing the Test subgroup (RG:DNI) with the Control subgroup (RG:DI600). Hence respondents in the Test subgroup and the Control subgroup are presenting on average, similar Dx values not only overall, as described above, but also for all Dx Factors. This is further indicating that the Dyslexia Index metric appears to be successfully identifying students with dyslexialike profiles from the research group of students with no reported dyslexia. Thus, as reported earlier, the foundation is laid for comparing the Academic Behavioural Confidence between the test and the control subgroups. As added verification, it can be seen that the converse outcome is established between the control subgroup and the base subgroup (ND400) where, with the exception of Dx Factor 3, Organization and Time Management, very highly significant differences between the Dx Factors means are recorded. This adds to the argument that overall, the base subgroup of students in research group ND  students who declared no dyslexic learning difference  who presented a Dyslexia Index of Dx < 400 can rightly be considered as presenting very low levels of dyslexianess and hence, the Dyslexia Index Profiler is presenting good discriminatory properties in accordance with its design rationale.
However for Dx Factor 3, Organization and Time Management, the mean Dx Factor values for all three research subgroups are not significantly different from each other which suggests that students at university who present very low levels of dyslexianess may be experiencing similar issues with organization and time management in their studies as do their dyslexiaidentified or not, peers. Mortimore & Crozier (2006), in their of students with dyslexia at university, draw on prior research (Gilroy & Miles, 1996, McLaughlin et al, 1994) to evidence the difficulties experienced by dyslexic students in organizing their study processes and timekeeping in addition to presenting their own results. Whilst the outcomes of their study were consistent with the earlier research drawn upon, their enquiry was conducted amongst students with dyslexia only and did not explore how the organization and timekeeping aspect of academic learning management may be referenced in comparison with students with no reported dyslexia. Hence the data summary presented here for this current project is an interesting outcome because firstly it is consistent with the findings of Mortimore & Crozier's study amongst students with dyslexia in that the mean Dx Factor 3 for the strongly dyslexic students in this project (RG:DI600) is at Dx = 635.53, but secondly because it has extended the level of enquiry to encompass students with no reported dyslexia, there arises the implication that on the basis of this metric's results and analysis, students who are strongly nondyslexic in other areas may be just as weak in organizational and timemanagement skills at university as students with dyslexia according to the indications of their outputs on the Dyslexia Index Profiler. Or viewing it another way, this is saying that most students at university tend to be disorganized and find timemanagement challenging and that this aspect of academic learning management is not unique to students with learning differences. A reaction to this might be that if universities are motivated to ensure that students studying on their courses become wellequipped to meet the learning challenges that they will be facing, then making early provision for upskilling students' organizational and time management competencies as part of the groundwork for enabling them to develop their academic learning management capabilities would be time wellspent and widely applicable across the student community. Adopting such a practice would also be inkeeping with the ethos of Universal Design for Learning (UDL) that has been consistently referred to in this thesis as an aspiration for studying at university.


Dx Factor 1 
Dx Factor 2 
Dx Factor 3 
Dx Factor 4 
Dx Factor 5 


Reading, Writing, Spelling 
Thinking & Processing 
Organization & Time Management 
Verbalizing & Scoping 
Working Memory 


∨ Research Subgroup: DNI Factor Means ∨ 
Dx Factor 
Research Subgroup: DI600
∨ Factor Means ∨ 
763.01 
647.60 
635.53 
734.64 
668.82 
Dx 1 
794.50 
p = 0.3325;
no significant difference (5%) 




Dx 2 
700.42 

p = 0.1449;
no significant difference (5%) 



Dx 3 
615.72 


p = 0.6302;
no significant difference (5%) 


Dx 4 
772.72 



p = 0.4828;
no significant difference (5%) 

Dx 5 
589.20 




p = 0.2254;
no significant difference (5%) 



Dx Factor 1 
Dx Factor 2 
Dx Factor 3 
Dx Factor 4 
Dx Factor 5 


Reading, Writing, Spelling 
Thinking & Processing 
Organization & Time Management 
Verbalizing & Scoping 
Working Memory 


∨ Research Subgroup: ND400 Factor Means ∨ 
Dx Factor 
Research Subgroup: DI600
∨ Factor Means ∨ 
276.43 
214.38 
586.78 
458.02 
377.68 
Dx 1 
794.50 
p < 0.0001;
very highly significant difference (0.1%) 




Dx 2 
700.42 

p < 0.0001;
very highly significant difference (0.1%) 



Dx 3 
615.72 


p = 0.3548;
no significant difference (5%) 


Dx 4 
772.72 



p < 0.0001;
very highly significant difference (0.1%) 

Dx 5 
589.20 




p < 0.0001;
very highly significant difference (0.1%) 



Dx Factor 1 
Dx Factor 2 
Dx Factor 3 
Dx Factor 4 
Dx Factor 5 


Reading, Writing, Spelling 
Thinking & Processing 
Organization & Time Management 
Verbalizing & Scoping 
Working Memory 


∨ Research Subgroup: ND400 Factor Means ∨ 
Dx Factor 
Research Subgroup: DNI
∨ Factor Means ∨ 
276.43 
214.38 
586.78 
458.02 
377.68 
Dx 1 
763.01 
p < 0.0001;
very highly significant difference (0.1%) 




Dx 2 
647.60 

p < 0.0001;
very highly significant difference (0.1%) 



Dx 3 
635.53 


p = 0.2550;
no significant difference (5%) 


Dx 4 
734.64 



p < 0.0001;
very highly significant difference (0.1%) 

Dx 5 
668.82 




p < 0.0001;
very highly significant difference (0.1%) 

An outcome of this analysis more generally and due to the increasingly demonstrable robustness of its construct validity, may be to suggest that the Dyslexia Index Profiler as it has been developed for this project as a discriminator as opposed to an identifier, is showing indications of merit in developing it as a either a screening tool for dyslexia  when dyslexia in higher education settings is framed in terms of parameters of academic learning management and studyskills  or better still, as a more widely available appraisal device in the toolkit for learning development/academic skills support services at university because it provides a readily comprehensible snapshot of any individual student's approach to study by generating a profile which identifies strengths that can be developed together with weaknesses that might be remediated.
Comparing differences in Dyslexia Index between research subgroups at a dimensional level
Further to examining differences in Dyslexia Index Factors, Dyslexia Index has been explored on a dimension by dimension basis as part of the process of trying to tease out which characteristics could be the more significantly responsible ones that might account for the differences in Academic Behavioural Confidence between the three research subgroups of particular interest (Test, Control and Base).
The summary table below list all, 20 dimensions of Dyslexia Index (Dx) and shows the mean Dx levels firstly between the two, core research groups  students who declared their dyslexia, and students who declared no dyslexic learning difference; and secondly between the three research subgroups  the Control subgroup, the Base subgroup and the Test subgroup. Data in the top section of the table shows the mean Dx values for each dimension  recall that the Dx value indicates the level of respondents' acquiescence with the dimension statement. Underneath, both the ttest pvalues and the Hedges 'g' effect size differences between pairs of groups and subgroups are shown. Aside from indicating the stark differences in the majority of the mean values between students with dyslexia and those without (RG:DI and RG:ND) when these data are further reduced into differences between the research subgroups, even more fascinating information is revealed. For example: looking at the differences in means for the TEST subgroup and the CONTROL subgroup for the two dimensions that together constitute the Dx Factor: Verbalizing and scoping it can be seen that the ttest output together with a mediumtolarge effect size indicate significant differences. This can be argued as evidence which supports the viewpoint that dyslexic students are likely to be academically more comfortable adopting planning strategies which permit a more holistic overview to be taken when approaching an assignment challenge rather then plan in lists or other linearthinking ways (Draffen et al, 2007)  hence the widely adopted feature of UK Disabled Students' Allowance provision of conceptmapping assistive technologies such as the applications 'Inspiration' and 'Mind Genius'. Both of these commonlyprovided software tools are designed to foster creative thinking, to facilitate ideasbrainstorming and patternspotting, and to enable the typical grasshopper thinking of the dyslexic student to be developed into meaningful learning from which powerful knowledge structures can be built, ordered and converted into a linear writing process (Novak & Canus, 2010). Evidence has also shown that conceptmapping applications as learning technologies as opposed to assistive technologies are gaining traction in curriculum design, both as an additional and accessible learning tool (Nesbit & Adescope, 2006), as a mechanism for summative assessment (Anghel et al, 2010) and not least in higher education contexts as a means to promote flexible learning approaches (Goldrick et al, 2014), all of which are the embodiment of UDL. Additionally, and of high relevance to students presenting weak spelling competencies, whether attributed to a dyslexia or not, is evidence from studies concerning TEFL (Teaching English as a Foreign Language) learners where conceptmapping applications have been very successfully used to develop Englishlanguage spelling skills by enabling spoken phonemes to be connected with their written forms in a highly innovative and relationshipbuilding format (AlJarf, 2011) and for connecting vocabulary to concepts in different contexts (Betancur & King, 2014).
Secondly, even more striking differences between dyslexic and nondyslexic students for verbalizing ideas in preference to writing about them (Dimension 4) are evidenced where the Dyslexia Index dimension mean value of Dx=84.34 for the strongly dyslexic students in the Control subgroup contrasts sharply with the mean value of Dx=42.34 for the strongly nondyslexic students in the Base subgroup. Indeed, it can be seen that for most of the dimensions, the differences in mean Dx values between dyslexic and nondyslexic students (RG:DI and RG:ND) are equally salient with the largest absolute difference being for Dimension 20, I get really anxious if I'm asked to read out loud, with a Dx Index difference of 32.57 points (RG:DI Dx=77.40, RG:ND Dx=44.83) which corresponded to an effect size of 0.9654, conventionally categorized as 'large'. For the corresponding difference in mean Dx values between the strongly dyslexic and strongly nondyslexic research subgroups, we see an even more substantial absolute difference of 62.11 Dx Index points (RG:DI600 Dx=83.38, RG:ND400 Dx=21.27) which is as we would expect given that dyslexia is conventionally considered as principally associated with reading difficulties. This considerable absolute difference in Dx points is echoed for Dimension 1, When I was learning to read at school I often felt I was slower than others in my class, where an even greater difference of 65.29 Dx Index points is recorded (RG:DI600 Dx=78.34, RG:ND400, Dx=13.05).
However an interesting difference which does emerge is observable from the data for Dimensions 2, My spelling is generally good (weak), 17, I get my 'lefts' and 'rights' easily mixed up and 10, In my writing at school, I often mixed up letter that looked similar, where the students with declared dyslexia in the Control subgroup present significantly higher Dx mean values than for their peers in the Test subgroup (Dx=75.45 / 49.17; Dx=75.28 / 57.78; Dx=67.17 / 45.33 respectively). For Dimension 2, spelling, this may be indicating that weak spelling is not a characteristic of students in the Test subgroup where in most other respects, these students are presenting dimensional levels of dyslexianess that are at similar levels to their identified dyslexic peers. This might be a reason to explain why students in the Test subgroup who might otherwise be considered as unidentified dyslexics have not had their dyslexic learning differences spotted at earlier stages in their learning careers as often poor spelling is the first indication that a learner may be dyslexic, especially to teachers, tutors or lecturers who are not wellversed in the range of other characteristics and attributes which frequently form part of the conventionally dyslexic profile.
Looking across the complete set of dyslexia dimensions, the outcomes that emerge when the Test and the Control subgroups are compared show that in all but 4 of the 20 dyslexia dimensions, the mean values for each of the dimensions respectively are very similar which is supported by generally small effect size differences and pvalues which indicate no significant differences between the means. This outcome is suggesting that the students in the Control subgroup who are presenting dyslexialike profiles are indeed dyslexic within the terms of reference of the Dyslexia Index Profiler. This is a pleasing outcome and adds to the construct validity of the Dyslexia Index metric that has been devised for this project as a mechanism for discriminating students who may be dyslexic amongst the research group of students who declared no dyslexia. Thus confidence is gained in using the measure as an index of a construct that is not directly observable (Weston & Rosenthall, 2003), in this case, termed 'dyslexianess', which is, of course, one of the foundations upon which this complete project is based. Smith (2005) summarizes the seminal work of Cronbach & Meehl (1955) on construct validity which comprehensively argues 'that the only way to determine whether a measure reflects a construct validly is to test whether scores on the measure conform to a theory, of which the target construct is a part' (op cit, p405) and it is argued that by exploring the contrasts in Dx Index values at a dimensional level and commenting on the extent to which the differences that have been measured are in keeping with the more widely accepted theoretical underpinnings of at least some of the typically observed characteristics of dyslexia, the construct validity of the Dyslexia Index Profiler is strengthened and justified as the discriminator for which it was designed in this project.
4.2 Factorial analysis of the Academic Behavioural Confidence Scale
The process of Principal Component Analysis has also been applied to the data collected on Academic Behavioural Confidence and likewise this has been conducted through the software application SPSS.
This has been prompted because a revised, 17item ABC Scale has been established by Sander & Sanders (2009) following a reinspection of the combined data from several of their earlier studies where ABC had been measured, and this was developed through PCA on data collected through their original 24item scale. Their (relatively) large composite dataset (n = 865) was established by aggregating data collected from five earlier studies conducted between 2001 and 2006 with psychology undergraduates from one university in South Wales (n=507), together with a further dataset of ABC values obtained from firstyear medical students at one HE institution in the Midlands collected in 2001 (n = 182) and an additional dataset of health care students from a new university in South Wales attending 6 different courses ranging from podiatry to dental health care (n = 176). The two, smaller of these three datasets are of a similar size to the number of respondents in this current project (n=166). There are both differences and similarities in the cohorts of students in Sander & Sanders' combined dataset when compared with feaures of my dataset. For example, one difference was that in the Sander & Sanders studies, students were all undergraduates whereas in my study students across the university community were invited to participate in the research, with the participation response producing an overall ratio between undergraduates and other students of 75% : 25%. ('undergraduates' includes students attending foundation or access courses and 'other students' comprises postgraduates, research students and a very small number of others who did not disclose their study level (3)). One similarity was that in Sander & Sanders' datasets, students' previous academic achievement at Alevel was recorded with the complete range of grades being presented and although this data was not requested in my study it is not unreasonable to suppose that students who responded to the invitation to participate presented an equally wide range of prior academic achievement. In the Sander & Sanders' datasets, students were drawn from a narrow range of subject specialisms whereas in my study, subjects studied at university was not recorded so it is also fair to assume that students from a range of curriculum specialisms are as likely to have participated as not.
The original Academic Confidence Scale (ACS) was formulated to operationalize an enquiry to explore stark differences in confidence observed between two very different student groups (Sander & Sanders 2003). The data collected was factoranalysed to reveal 6 subscales: Studying, Understanding, Attendance, Grades, Verbalizing and Clarifying although it was pointed out that this resulting factor structure was a bestcompromise as some statements in the ACS did not load on to only one factor. The ACS was renamed shortly after its inception to the Academic Behavioural Confidence Scale to acknowledge that the scale was in fact more sharply focused on measuring students' confidence in actions and plans related to academic study (Sander & Sanders, 2007). The later, factor analysis of the aggregated data demonstrated that this revised scale consisted of 6 factors: Grades, Studying, Verbalising, Attendance, Understanding and Requesting which was deemed a better representation of the subscale structure than the earlier 6factor analysis as a result of a more detailed, Confirmatory Factor Analysis. However an additional outcome of the factor analysis was to identify some scaleitem redundancy which led to 7 scale items being removed from the original, 24item ABC Scale. A further factor analysis was then conducted which revealed a new factor structure with only 4 factors: Grades, Verbalizing, Studying and Attendance.
The data collected in this project has been acquired through the complete, original 24scaleitem scale and since Sander & Sanders' 17point scale was revised by merely omitted some scale items leaving the remainder exactly as they had been in the earlier, 24item scale, it has enabled both ABC24 and ABC17 outputs to be generated from my local data. The table below presents these outputs for comparison:
Research Group 
Research subgroup 
n 
ABC24 mean 
ABC24 sd 
ABC17 mean 
ABC17 sd 







DI 
DI600 (Dx > 592.5) 
47 
57.89 
15.24 
57.49 
15.75 
Hedges' g effect size / Student's ttest pvalue: DI600:DNI: 

g = 0.483 / p = 0.041 

g = 0.521 / p = 0.032 

ND 
DNI (Dx > 592.5) 
18 
64.92 
12.43 
65.24 
12.26 








ND400(Dx < 400) 
44 
72.15 
12.35 
72.25 
12.66 
This reveals little difference between the mean ABC24 and mean ABC17 values for any of the research subgroups, showing that a slightly greater effect size is generated using the 17point ABC Scale, this being between the sample means for research subgroups DI600 and DNI. In both cases (ABC24 and ABC17) Student's ttest reveals that a significant difference (p < 0.05) is present between the sample means (onetail test, 5% level).
However, it is unusual to be able to use two, so closely related versions of a metric to evaluate the same construct and reflecting on this identified the possibility of applying a statistical test to evaluate whether or not there is a significant difference in the effect size outputs that the two versions have generated for this local data. By doing so, were this to produce a result that indicated no significant difference between the effect sizes, this would imply that taking either verion of the metric would be simply a matter of preference. To date, no literature has been found where this idea has been converted into a workable statistical process that might be followed as an exemplar or to offer guidance about how a comparison of effect sizes in this context might be achievable, especially since the distribution of effect sizes is unknown. But it is possible to calculate a confidence interval for the population Cohen's 'd' effect size, 'δ', this being an effect size measure of which Hedges 'g' is a slightly more refined version, from which it may be possible to establish whether these two effect sizes are in fact (statistically at least) the same. The Confidence Interval calculation process for Cohen's 'd' that is accessible (Cumming, 2012) generates the confidence interval for the estimated population effect size and when using the ABC24 Cohen's 'd' effect size difference between the ABC mean values between research subgroups DI600 and DNI came out at 0.068 < δ < 1.032, and for the ABC17 mean values, 0.032 < δ < 1.070. Given the very close match between these two confidence intervals, and in the absence of a process to generate an independent samples difference between the means using the conventional Student's ttest process, this is suggesting that to all intents and purposes, the difference in effect sizes when using ABC24 compared with using ABC17 is marginal. Sander & Sanders (S&S) claim that the criterion validity of the ABC Scale is enhanced through their factor analysis procedure and the subsequent reduction into a 17point scale.
Much has been drawn from the statistical rigour that Sander & Sanders have demonstrated to justify the robustness of their Academic Behavioural Confidence Scale and given the advancing reputation that the metric is gaining as a wellproven and valid scale for exploring academic confidence amongst university students, it is being used in this project without hesitation as the best metric available for exploring the issues being considered. However in the interests of trying hard to ensure that analysis of the ABC Scale's output is scrutinized very carefully and contextually in respect of the datapool that has generated the results, there are reasonable grounds for conducting PCA on this project's local data to determine whether a similar or different factor structure emerges which can be more acutely integrated with the datapool's output for the Dyslexia Index Profiler. Additionally, the possibility should be considered that there may be an unwitting bias in Sander & Sanders' analysis due to students being all undergraduates and all from similar subject specialisms whereas in this project, no data was collected about students' subject specialisms nor their levels of study. Hence it is possible that this may impact on the applicability of exploring outputs from analysis of the subscales of the existing ABC24 Scale's 6factor, or the ABC17 Scale's 4factor structure with data collected from a more generalized source. Recall that earlier attention has been drawn to the reduced effectiveness of a metric developed from a closed cohort sample from a single university when used to explore the same construct as presented in a sample taken from a different university's student community (the YAA Adult Dyslexia Scale; (Hatcher & Snowling, 2002), adapted for use in an Australian university with disappointing results (Chanock et al, 2010)). Hence the specific factor structure that is revealed through PCA may be more appropriate to use in this project's comparative analysis with Dyslexia Index as opposed to relying on the factor structure and revised 17item ABC scale determined from the Sander & Sanders' collection of studies. It is reasonable to suppose that Corkery (2011) followed a similar line of reasoning to justify applying principal component analysis to the local data in her study and indeed, the factors which emerged showed differences in comparison to both the Sander & Sanders (S&S) factor structures for both the 24scaleitem and the 17scaleitem ABC Scales. This at least sets a precedent for this approach to applying PCA to local data within a research project when examining the outputs from the application of the Academic Behavioural Confidence Scale. However, this may raise an issue about the stability of the ABC factors and suggests that researchers choosing to use the metric in their studies may be wise to explore the factor structre of the ABC Scale in relation to their local data unless it could be shown that the demographics of their research cohorts may closely resemble those of Sander & Sanders' original (combined) studies. On a more general point, but also in line with Chanock's (2010) observations in respect of the York Adult Assessment (for dyslexia) about the inadvisability of using evaluatative metrics developed entirely from one single institution's local datapool in a different institution, this suggests that research outputs may be more robust were researchers more inclined to extend their data analysis procedures to include principal component analysis on their local data when using metrics developed elsewhere.
Hence the Principal Component Analysis applied to the 24scaleitem Academic Behavioural Confidence Scale used to acquire this project's data has resulted in 5 factors being identified. However, just as for the factorial analysis conducted on the Dyslexia Index scale, it can been seen from the Rotated Component Matrix (below) that the factor structure that emerged for this analysis of the ABC Scale was not as simple as is desirable because a few dimensions (that is, ABC scale items) loaded on to more than one factor. Hence once again, by applying an element of best reasonable judgement, it is considered that there is justification for accepting the outcomes and in accordance with the 'type' or 'sense' of scale items that emerged as sensibly loading onto each of the 5 factors, they have been categorized as:
 ABC24 Factor 1:  Study Efficacy
 ABC24 Factor 2:  Engagement
 ABC24 Factor 3:  Academic Output
 ABC24 Factor 4:  Attendance
 ABC24 Factor 5:  Debating
Rotated Component Matrix for Academic Behavioural Confidence 24point scale 
item # 
item statement 
Factor 
Communalities 
ABC 

1 
2 
3 
4 
5 
Extraction 


study efficacy 
engagement 
academic output 
attendance 
debating 

121 
 plan appropriate revision schedules 
0.809 




0.761 
101 
 study effectively in independent study 
0.703 




0.637 
104 
 manage workload to meet deadlines 
0.695 




0.593 
113 
 prepare thoroughly for tutorials 
0.665 




0.578 
122 
 remain adequately motivated throughout my time at university 
0.639 




0.555 
119 
 make the most of university study opportunities 
0.637 




0.570 
114 
 read recommended background material 
0.602 


0.318 

0.530 
103 
 respond to lecturers' questions in a full lecture theatre 

0.799 



0.662 
110 
 ask lecturers questions during a lecture 

0.774 



0.707 
112 
 follow themes and debates in lectures 

0.654 



0.610 
105 
 present to a small group of peers 

0.624 



0.483 
102 
 produce your best work in exams 

0.605 


0.444 
0.692 
111 
 understand material discussed with lecturers 

0.597 



0.516 
117 
 ask for help if you don't understand 

0.454 



0.406 
116 
 write in an appropriate style 


0.819 


0.736 
115 
 produce coursework at the required standard 


0.814 


0.805 
107 
 attain good grades 
0.383 

0.740 


0.740 
120 
 pass assessments at the first attempt 


0.696 


0.593 
123 
 produce best work in coursework assignments 
0.492 

0.511 

0.344 
0.649 
106 
 attend most taught sessions 



0.812 

0.739 
124 
 attend tutorials 



0.772 

0.675 
118 
 be on time for lectures 



0.676 

0.522 
108 
 debate academically with peers 

0.435 


0.640 
0.652 
109 
 ask lecturers questions in oneone settings 
0.321 
0.346 


0.632 
0.624 








It is useful to consider this factor analysis in the light of that of Sander & Sanders (2009) study, reproduced below. An attempts has been made to compare the grouping of dimensions into factors that emerged from their PCA to the PCA on this project's local data, indicated by what has been termed the 'closest map'. This is where dimensions from both the S&S PCA and my own PCA result in similar dimensional groupings. It has been necessary of course to revert back to Sander & Sanders original 24item scale to make this comparison. It can been seen that the first four factors of both analyses present similar loadings and groupings of dimensions although Sander & Sanders' factors 5 and 6 draw no obvious mapping to the factor that have emerged out of my PCA.
Rotated Component Matrix for Academic Behavioural Confidence 24point scale
(adapted from: Sander & Sanders, 2009, p25) 
item # 
item statement 
Factor 
ABC 

1 
2 
3 
4 
5 
6 

Sander & Sanders' factor designations: 
studying 
verbalising 
grades 
attendance 
understanding 
requesting 

closest map to ABC24(5) in my data: 
study efficacy 
engagement 
academic output 
attendance 
no mapping 
no mapping 
121 
 plan appropriate revision schedules 
0.80 





101 
 study effectively in independent study 
0.72 





122 
 remain adequately motivated throughout my time at university 
0.62 





104 
 manage workload to meet deadlines 
0.56 





103 
 respond to lecturers' questions in a full lecture theatre 

0.85 




105 
 present to a small group of peers 

0.81 




108 
 debate academically with peers 

0.67 




110 
 ask lecturers questions during a lecture 

0.58 




120 
 pass assessment at the first attempt 


0.83 



115 
 produce coursework at the required standard 


0.74 



116 
 write in an appropriate style 


0.67 



107 
 attain good grades 


0.66 



123 
 produce best work in coursework assignments 


0.55 



102 
 produce best work in exams 


0.51 



124 
 attend tutorials 



0.86 


106 
 attend most taught sessions 



0.82 


118 
 be on time for lectures 



0.40 


119 
 make the most of university study opportunities 
0.17 
0.24 

0.29 

0.21 
113 
 prepare thoroughly for tutorials 




0.73 

112 
 follow themes and debates in lectures 




0.72 

111 
 understand material discussed with lecturers 




0.68 

114 
 read recommended background material 




0.68 

109 
 ask lecturers questions in oneone settings 





0.85 
117 
 ask for help if you don't understand 





0.83 








It is of note that scale item 119 was not attributed to any of Sander & Sanders' factors with the highest loading of just 0.29 with the factor 'attendance'.
A cursory inspection of the two tables sidebyside shows that:
 My Factor 1, 'study efficacy' includes all four dimensions in S&S Factor 1: 'studying', with two of the remaining 3 dimensions attributed into S&S Factor 4, 'understanding' and the final dimension, 119, 'make the most of university study opportunities' being unattributed in the S&S analysis; however with a loading factor of 0.637 in my data, this dimension should clearly be included in my Factor 1. Where Sander & Sanders designate these dimensions into two factors 'studying' and 'understanding', these two factors together map to my Factor 1, so aspects of academic confidence at university that Sander & Sanders call 'studying and understanding', I call 'study efficacy'. This matches well.
 3 of S&S's 4 dimensions in their Factor 2, 'verbalising' map to to the same 3 out of 7 dimensions in my Factor 2, 'engagement'. S&S include dimension 108 'debate academically with peers' into their Factor 2 whereas I attribute this dimension to my Factor 5, 'debating'. Two further dimensions in my Factor 2 are attributed in the S&S analysis to their Factor 4, 'understanding' so this is suggesting that the S&S Factors 'verbalising' and 'understanding' when taken together make a close map fo my Factor 2 'engagement'.
 In S&S Factor 3, 'grades', 5 out of the 6 dimensions in this factor also appear in my Factor 3, 'academic output' so this is a close mapping between the two factors. The additional dimension, 'produce best work in exams' in S&S Factor 3 however, presented a higher loading with Factor 2, 'engagement' in my analysis.
 My Factor 4, 'attendance' contains exactly the same 3 dimensions as S&S Factor 4, 'attendance' so there is an exact mapping here.
 S&S designated a Factor 6, 'requesting' which contained dimensions 109 and 117, 'ask lecturers questions in oneone settings' and 'ask for help if you don't understand' whereas the former of these (109) is grouped with dimension 108, 'debate academically with peers' in my PCA with these two dimensions alone forming the final Factor 5, 'debating' in my analysis, so this acknowledges more of a dialogical interactional relationship between not only students and their peers, but also with their academic staff.
The extraction commonalities was not published in Sander & Sanders 2009 paper from which this data has been drawn.
4.3 Dx Factor x ABC Factor Matrix
The principal component analysis applied to the data acquired through the Dyslexia Index metric suggests that it loads onto 5 factors and these have been designated:
 Reading, Writing, Spelling
 Thinking and Processing
 Organization and Timemanagement
 Verbalizing and Scoping
 Working Memory
and that the PCA applied to data collected on Sander & Sanders full, 24item Academic Behavioural Confidence Scale has also loaded onto 5 factors, designated:
 Study efficacy
 Engagement
 Academic output
 Attendance
 Debating
To explore the interrelationships between each of these two sets of 5 factors, a 25 x 25 cell matrix has been constructed (below) which sets out Hedges' 'g' effect size and Student's ttest pvalues between the three, key research subgroups, DNI (TEST), DI600 (CONTROL) and ND400 (BASE) when these are reestablished according to each of the Dyslexia Index factors. It is cruicially important to understand that the process of resifting respondents' datasets according to new criteria set by Dx Factors results in different cohorts of students comprising each of the key research subgroups when the Dx boundary value of Dx = 592.5 is freshly applied to the core, research groups, ND and DI when these are reordered according to the five Dyslexia Index factors respectively.
Thus, sequentially ordering and reordering the complete datapool into research groups and subgroups according to Dx values for each Dx factor has generated a fresh opportunity to analyse the data from a deeper perspective. The objective is to see if this can lead to a better understanding about the impacts of specific groups of dyslexia dimensions on academic confidence overall, and also on the components of academic behavioural confidence where these too, have been established by the process of principal component analysis. Hence, the 5factor x 5factor matrix of interrelationships has been constructed.
To understand more clearly the process of dataset ordering/reordering according to Dx Factor, consider student respondent #96408084 in the datapool who is in research group ND because no reported dyslexia was declared but who presented an overall Dyslexia Index of Dx = 604.94. At the outset, this Dx value placed this respondent into the TEST research subgroup, DNI, which is students with an unreported dyslexialike profile and as with the complete datapool, together with the initial declaration of dyslexia/no dyslexia, applying the Dx boundary fence has sorted all datasets into groups and subgroups accordingly which has enabled an overall idea to have been established about the impact of dyslexia on academic confidence, summarized in Key Outcomes 1 and 2 above.
However when this student's dataset is reviewed on a factor by factor basis, the Dx factor values are quite disparate ranging from Dx = 824.11 in Factor 1, Reading, Writing, Spelling, to Dx = 80 in Factor 4, Verbalizing and Scoping. So although this respondent is sifted into research subgroup DNI on the basis of overall Dx value, because at Dx = 604.94 this is above the boundary value of Dx > 592.5, when the Dx value for each factor is considered in turn this student will be sifted into research subgroup DNI only for Factors 1 and 2, as Dx values for the remaining factors are below the boundary value of Dx = 592.5. Indeed with a Dx Factor 4 value of Dx = 80, this particular student would actually feature in research subgroup ND400 for Dx Factor 4 Verbalizing and Scoping, which is the research subgroup of students presenting very low levels of dyslexianess, and for Dx Factors 3 and 5, this student's data would not be sifted into any of the three, key research subgroups on which values of Academic Behavioural Confidence are being compared.
Dx overall 
Student respondent 
Dx Factor 1 
Dx Factor 2 
Dx Factor 3 
Dx Factor 4 
Dx Factor 5 


Reading, Writing, Spelling 
Thinking & Processing 
Organization & Time Management 
Verbalizing & Scoping 
Working Memory 







604.94 
#96408048 
824.11 
746.99 
512.26 
80.00 
489.51 
= RG:DNI (TEST) 

= RG:DNI (TEST) 
=RG:DNI (TEST) 
=RG:NDx400 
=RG:ND400 (BASE) 
=RG:NDx400 







Taking this individual as a single example, the significant point is that the Dx values for Dx Factors 1 and 2 are high, suggesting that this particular student is presenting a strongly dyslexic profile in the two (factor) areas of Reading, Writing, Spelling, and Thinking & Processing – conventionally regarded throughout decades of dyslexia research with children as being key indicators of the syndrome. Thus given that no declaration of dyslexia was indicated on this student's questionnaire response, and assuming this was not a falsity, it may be reasonable to suppose that according to the conventional assessment criteria at least, this individual is presenting a dyslexic learning difference which so far, has not been detected. However, on other factor measures, this student does not present other typical dyslexic characteristics quite so strongly and hence on balance, given a more conventional dyslexia screening, this student may not have been identified as dyslexic.
It must be reemphasized that the datasets which comprise research subgroups ND400, DNI and DI600 after sifting the datapool according to Dyslexia Factor 1 values for example, will or may contain a different collection of datasets when the sifting criteria is according to Dx Factors 2, 3, 4 and 5. The table below summarizes how the sizes of the research subgroups change as a result of this process. These redistributions of the datapool enable a alternative insight to be taken into differences in academic confidence (at a factoral level) between students with reported dyslexia and those with unreported dyslexialike profiles on a Dx factorbyfactor basis. At the outset this appears to be overcomplicating the analysis process but it is justified because the results that emerge appear to be showing a range of significant differences across the research subgroups when these are examined with Dyslexia Index Factor as the determining criteria. As reported in the literature review earlier, examining dyslexia at a factorial level has been gaining traction in recent research and particularly a late study conducted with students at the University of Amsterdam (n=154) demonstrated nine distinguishable factors of dyslexia. These were classified as: Spelling, Reading, Rapid Naming, Attention, ShortTerm Memory, Confusion, Phonology, Complexity, and Learning English (Tamboer et al, 2017). Note the researchers' comment that this final factor, Learning English, might only have become significant in their cohort of students because these individuals were highly academically competent in their own language (Dutch) and likely to have overcome many of their earlier linguistic difficulties in their native tongue, but not with foreign languages. It may be reasonable to discount this factor in student coherts with L1 English. This paper has been particularly singled out not only as an example of a study that has explored dyslexia at a factorial level, but also because the cohort of research participants used in the study closely resemble those in my project, particularly because they comprised known dyslexic students, those who were clearly presenting no indications of dyslexia as determined by any of the conventional criteria, but also, a significant subgroup of maybedyslexics emerged out of the analysis. The outcomes of the study have a bearing on the factorial analysis outcomes in my project due to the similarities of the factors and the process by which they were established. This is despite Tamboer et al's principle interest being determination of the predictive validity of their newlydeveloped screening test for dyslexia in the Dutch language. Their research suggested that this validity was strong, leading to the conclusion that the screening test that had been developed would be useful as a dyslexia identifier in higher education contexts. Of particular interest was that the selfreport questions that had been included in their data collection instrument also returned high construct validity and significantly, an even higher predictive validity than the other tests that had been included in the screener (ibid, p167). To date, no studies have been found which use a factorial analysis of a dyslexia evaluator in higher education settings as an independent variable correlator for exploring another construct, in the case of this project, academic confidence.
Reconstructed Research Subgroups
Thus, the outcomes of the process described directly above is shown in the table below which summarizes the relative sample sizes of the reconstructed research subgroups when these are sifted according to Dyslexia Index Factor using the established boundary value of Dx=592.5. Recall that this boundary value provided the criteria for sorting students from research group ND into the TEST subgroup: RG:DNI, and sorting students from research group DI into the CONTROL subgroup: RG:DI600. By determining through the ttest that the mean Dx values were not significantly different between the TEST and the CONTROL subgroups when this boundary Dx value was employed it has been argued that values of Academic Behavioural Confidence can therefore be properly compared because these two subgroups are presenting (statistically) the same mean Dyslexia Index.
Thus it has been important to run the ttest comparator again when datasets have been sifted into reconstructed research subgroups on a Dyslexia Index factorbyfactor basis because the datasets comprising each research subgroup are reconstituted 5 different ways according to each of the 5 Dx factors and consequently, the mean Dx values are not consistent. Therefore it has been important to ensure that the Dyslexia Index means remain not significantly different across the research subgroups. The ttest evidence shows that these sample means can be considered as not significantly different between the two reconstructed research subgroups of interest (DNI and DI600). It is therefore appropriate to apply the earlier rationale, that is is justifiable to consider ABC effect sizes on an ABC factorbyfactor basis and to look for significant differences between ABC mean values all between the key research subgroups DNI and DI600 as these become reconstructed on a Dx factorbyfactor basis. [Note: the designation 'Dx20(5)' is reminding us that the 20scaleitem Dyslexia Index Profiler with 5 factors is the data source.]
This has led to the construction of the Dx Factor x ABC Factor Matrix (below) which is the most important data analysis outcome of the complete project.
It presents effect size differences and ttest outputs between mean Academic Behavioral Confidence values for the test research subgroup, DNI, the control research subgroup, DI600, and the base research subgroup, ND400 when these research subgroups are reconstructed on the Dyslexia Index factorbyfactor basis. Two sets of comparators are considered to be pertinent: firstly between the group of students considered to be highly NONdyslexic and those others considered to be highly dyslexic (RG:ND400 and RG:DI600); and secondly between the control research subgroup (RG:DI600) and the test subgroup (DNI) and these data are presented in rowpairs for each Dyslexia Index Factor. In addition to the effect size differences being shown, the absolute ABC values are provided to contextualize the effect size values and the overall key findings of the analysis which relate back to the research hypotheses are indicated in the bottomright of the matrix (the red box). To aid clarity, pvalues for differences between factor mean ABC values for research subgroups ND400 and DI600 have been omitted with the exception of the overall result (bottomright). However, where these are apposite, these data are provided in the discussion which follows below.
For example: consider the row of data for Dyslexia Index Factor 3: Organization and Time Management. When research subgroup DNI (from research group ND) and research subgroup DI600 (from research group DI) are reconstructed using Dx Factor 3 as the sifting criteria (n=49, n=35 respectively, from the table of Sample Sizes above), it can be seen that the mean average for ABC Factor 1: Study Efficacy (for example) for the respondents in these reconstructed subgroups shows an effect size of 0.42 supported by a significant difference between the ABC Factor 1 sample means (p=0.0299). The represented the statistical analysis of the differences between the mean ABC values of ABC=65.3 and ABC=57.0 for research subgroups DNI and DI600 respectively. In other words, students with declared dyslexia present Academic Behavioural Confidence that is significantly lower, statistically, than students in the datapool with no declared dyslexia but who present dyslexialike profiles, when the data is arranged according to the Dx Factor, Organization and Time Management.
Hedges 'g' has been used because this calculation uses a weighted mean process for pooling the standard deviations of each dataset being considered which is important when the datasets are of different sizes; Student's ttest for independent sample means is used in the onetail format because in almost all cases, the mean ABC24 values for the (test) research subgroup DNI exceeded those for the (control) research subgroup DI600.
Effect size between Academic Behavioural Confidence:
The matrix of effect sizes above is the most substantial data analysis outcome of the research because as well as highlighting the overall effect size between the Academic Behavioral Confidence values that address the main research hypotheses, it has also identified significant effect sizes in ABC factors that may not have otherwise been revealed.
The central analysis outcome evidences a very highly significant difference between the mean values of Academic Behavioural Confidence between students with identified and declared dyslexia (RG: DI) and students who declared no known, dyslexic learning challenges.
Key Outcome 1:
KEY OUTCOME 1:
On the basis of the analysis of this datapool, students with no indications of dyslexia are shown to have, on average, a substantially higher level of Academic Behavioural Confidence than their dyslexiaidentified peers.
[RG:ND400, ABC=72.3; RG:DI600, ABC=57.9; Effect size: g=1.04; Ttest: t=5.037 , p<0.0001]
The second, important overall key finding, which is of the essence of this project, is that the analysis identifies a medium, Academic Behavioural Confidence effect size of 0.48 between the key research subgroups: DNI and DI600 (shown in the gridsector extreme bottomright). This is the effect size difference between the mean ABC values of research subgroups that are constructed on the basis of the overall Dx values of respondents. Taken together with the ttest for significant difference between independent sample means which returns a pvalue of p=0.043 (t=1.743, onetail test) these outcomes are indicating a significant difference between the ABC sample means for these research subgroups. DNI and DI600 which address one of the enquiry hypotheses and enables an important statement to be made:
Key Outcome 2:
KEY OUTCOME 2:
On the basis of the analysis of this datapool, students identified with unreported, dyslexialike profiles are shown to have, on average, a significantly higher level of Academic Behavioural Confidence than their dyslexiaidentified peers.
[RG:DNI, ABC=64.9; RG:DI600, ABC=57.9; Effect size: g=0.48; Ttest: t=1.743 , p=0.043]
These first two key outcomes directly address the project's core, research hypotheses: firstly that students with dyslexia present a significantly lower academic confidence, as measured through the Academic Behavioural Confidence Scale, and likely, a lower level of academic selfefficacy, than their nondyslexic peers. Key Outcome 1 evidences this conclusion. Secondly, Key Outcome 2 appears to be indicating that students with unreported dyslexia, using this phrase tentatively because the Dyslexia Index Profiler has been constructed so as not to be a screening tool for dyslexia per se, present a measurably and significantly higher level of academic confidence than their equivalently dyslexic peers, as determined through the ABC Scale, and also as likely, a higher level of academic selfefficacy. These are highly significant results because they may be suggesting that students at university with dyslexia but who may be unaware that they may be dyslexic are best left incognizant of the fact because to otherwise suggest that they consider taking a dyslexia screening test which may subsequently demonstrate that they have a dyslexic learning difference might adversely affect their academic confidence, their academic selfefficacy and perhaps their academic achievement.
Exploring Academic Behavioural Confidence on a Dyslexia Index factorbyfactor basis
By looking in more detail at the matrix of effect size and pvalue results for the component analysis for both metrics (Dx and ABC) it may be possible to identify where the contributing differences between Dx and ABC for each of the subgroups lies. In the first instance this examination is conducted on a Dyslexia Index factorbyfactor basis, following which a short summary discussion distills the outcomes and attempts to understand what they mean.
Dyslexia Index Factor 1: Reading, writing, spelling
Stemming from the earliest observations by PringleMorgan (1896) and Hinshelwood (1896) about individuals afflicted with a wordblindness, the overwhelming body of research about dyslexia takes persistent reading difficulties and the retarded development of reading skills in youngsters in comparison with their peers as the primary indicative factor of a dyslexia. In the earlier decades, published studies where dyslexia had been used as a term in conjunction with reading disability/disabilities or reading difficulty/difficulties appear scarce with a literature search conducted with through GoogleScholar returning just 39 studies.
Dyslexia Index Factor 2: Thinking and processing
content
Dyslexia Index Factor 3: Organization and time management
content
For example, for respondents organized according to Dyslexia Index Factor 3: Organization and Time Management, this is then the only Dx factor ‘sift’ that presents notable effect size differences between the research subgroups DNI and DI600 data in all five factors of Academic Behavioural Confidence. Effect size ‘g’ values range from g = 0.38 in ABC factor 5: Debating with the ttest indicating an albeit only just significant difference between the sample means (p= 0.046); to an effect size of g = 0.89 in ABC factor 2: Engagement. The ttest returned a very highly significant pvalue of p=0.0001 (rounded to 4 dp, the actual pvalue is p = 0.0000569). Given that effect size differences are effectively ‘one tail’, that is, are set so that a positive effect size indicates that ABC is higher for the research subgroup DNI than subgroup DI600, these results seem to be indicating that students with reported dyslexia exhibit significantly lower levels of academic confidence when sifted according to their Organization & Time Management factor of Dyslexia Index. This might be suggesting that on the basis of this dyslexiaindicating factor at least, aspects of dyslexia support related to ameliorating apparent weaknesses in organization and time management may be less effective than might be supposed. Not knowing that you may be dyslexic appears to be better for you when it comes down to the studyskill attribute of organization and time management.
It is also highly interesting to note that for this Dx Factor 3, the effect size differences between students regarded as highly NONdyslexic (that is, research subgroup ND400) and the dyslexic control group are all negative. I think this is therefore demonstrating that when considering a level of dyslexia as measured through the parameter, Organization and Time Management, it is better to be a student with an unreported dyslexialike profile than it is to be either a reported dyslexic or highly nondyslexic. This is puzzling but may be indicating that very curiously, some the dimensions of dyslexia that constitute this factor are actually positive attributes in relation to academic confidence but only in students with (potentially) unidentified dyslexia. Clearly conclusions are in relation to this datapool of respondents and hence caution is required in any attempt to generalize more widely, even though research subgroup DNI for this Dx Factor is quite sizeable (n=49) in relation to the total size of the datapool (n=166).
Dyslexia Index Factor 4: Verbalizing and scoping
content
Dyslexia Index Factor 5: Working memory
content
[NOTE: move the following paras to incorporate with the Dx factor sections]
It must be emphasized again that the Dyslexia Index factor analysis process used here does generate different cohorts of students in each research subgroup when regarding Dyslexia Index (Dx) as the independent variable – that is, the one I’ve fixed or chosen. This is because the process of considering the aggregate of the values for each of the dimensions that together constitute a factor generates a different Dyslexia Index than it might for any other factor for any specific student respondent. In other words, Student X will have a different Dx value for each Dx factor which will be different from their overall (i.e. aggregated) Dyslexia Index, and this may mean that the student is included or not in any of the research subgroups of interest, ND400, DNI, and DI600 on the basis of that factor, where the same student may be included or not, when generating a Dx value through one of the other Dx factors. [Perhaps I should build fresh diagramatic visualizations for students to show the different Dx values they present against each Dx factor.]
This point is demonstrated here:
For example, consider respondent #96408048 from research group ND who presented an overall Dyslexia Index of Dx = 604.94, hence placing this respondent just above the boundary into research subgroup: DNI – that is, students with an unreported dyslexialike profile. The Dyslexia Index values for each of the 5 factors of Dyslexia Index for this respondents are these:
Dx overall 
Student respondent 
Dx Factor 1 
Dx Factor 2 
Dx Factor 3 
Dx Factor 4 
Dx Factor 5 


Reading, Writing, Spelling 
Thinking & Processing 
Organization & Time Management 
Verbalizing & Scoping 
Working Memory 







604.94 
#96408048 
824.11 
746.99 
512.26 
80.00 
489.51 







The factor analysis reveals that this respondent's Dyslexia Index is greater than the subgroup boundary value of Dx = 592.5 for only two of the factors. What it is interesting to note is that this respondent's Dx values for those two factors is high, indicating that this particular individual is presenting a strongly dyslexic profile in these two areas  reading, writing, spelling, and thinking & processing  conventionally regarded throughout decades of dyslexia research with children as being key indicators of the syndrome. Reflecting on this has caused me to consider the ways in which the factor Dx values are contributing to the overall Dx value and additionally, how the factor profiles of the other respondents in research subgroup DNI (sifted according to the overall Dyslexia Index value of Dx > 592.5) compare to each other.
Although the matrix highlights many interesting feature and differences which will be discussed further below, it is significant to note that when the datapool is sifted according to Dx Factor 1: Reading, writing , spelling, the differences in academic behavioural confidence between students with dyslexia and those with possible unidentified dyslexia, (RG:DI600 and RG:DNI) are small and statistically not significant for neither the overall ABC value nor for any of the ABC factor values. This may be suggesting that where students in each of these research subgroups respectively may be presenting similar levels of dyslexiarelated issues in the context of literacy skills, these do not impact significantly on levels of academic behavioural confidence, whereas for the other four Dx Factors, differences in ABC are more pronounced. This could be consistent with arguments which support the view that for many dyslexic individuals operating at the higher levels of academic capability required at university, earlier literacy difficulties associated with their dyslexia which may have been apparent in preuniversity learning have responded to the development of strategies and compensations that have made literacy challenges less problematic.
[14593 / #]