Exploring The Relationship Between Impulsivity And General Anxiety In ADHD Students

Background

The objective of the given study is to predict if impulsivity symptoms can be used as a reliable estimator of general anxiety in case of ADHD patients. The given study is relevant since exploring such a relationship would help in understanding the extent of impact in academic performance of ADHD students based on the underlying measure of impulsivity.

Primary data has been collected for this study. The relevant sample comprises of 473 college students who have been diagnosed with ADHD along with a control group which comprises of non-ADHD suffering college students whose count is 204. It is noteworthy that the control sample size was later reduced to 200 since 4 students who had previously been diagnosed with ADHD were excluded from the group. Further, in selection of sample care has been taken to match the control group with the affected group with regards to year in school and ethnicity so as to minimise the impact of these variables.  The DSM-IV-TR symptoms of generalized anxiety and panic attacks related form was filled by both the affected group and control group which provided the requisite data (Provatt et. al, 2015). The code (-9) has been used for highlighting missing data.

Hypothesis

The given study expects to find that impulsivity does tend to be a significant input variable for prediction of general anxiety levels. This would require that the slope of the regression model using impulsivity as the independent variable and general anxiety as the dependent variable would be significant and hence it cannot be assumed to be zero.  However, failure to reject the null hypothesis would imply an insignificant relationship between the variables of interest.

Data Description

The given data has a host of variables but two variables are essentially significant from the perspective of the current study. One of these measures impulsivity and ahs been labelled as HYPER. On the other hand, another variable named GADD tends to measure the general anxiety level. There are other variables related to age, gender, ethnicity, academic performance, year which have also been provided but have limited relevance. The group of students with ADHD are part of group 1 while group 2 is essentially the control group.  Also, it is noteworthy that for analysis only those students have been considered which do not have any missing data in HYPER and GADD. Further, both groups have been included in the analysis (Provatt et. al., 2015).

Results

The relevant scatterplot between HYPER (Independent Variable) and GADD (Dependent variable) is indicated below.

The above graph clearly suggests that the relationship between the two variables is not linear. It is essentially in a form of a round blob where for same values of Hyper variable, differing values of GADD are visible. Hence, it might be advisable to transform the data into log to obtain a better fit (Flick, 2015). However, for the purposes of this study, no transformation has been performed. Further, there are four outliers present in the data which have been identified using marker tool. Three of these have HYPER value exceeding 30. These four points despite being outliers do not have a significant impact in the correlation analysis considering that the remainder pattern also does not suggest any linear or significant relation between the given variables. The correlation coefficient for the given variables has been computed as follows.

Correlations
	GADD	Hyper
GADD	Pearson Correlation	1	.248**
Sig. (2-tailed)		.000
N	349	349
Hyper	Pearson Correlation	.248**	1
Sig. (2-tailed)	.000
N	349	350
Correlation is significant at the 0.01 level (2-tailed).

Hypothesis

From the above computation, it is apparent that the sample size is actually much reduced from the initially collected data owing to the missing data. Further, the correlation coefficient is 0.248 which implies that even though correlation is weak but it is significant as indicated from the above output (Hillier, 2016).

The linear regression model is obtained with HYPER acting as the independent variable and GADD acting as the dependent variables. The relevant screenshot is indicated below.

Model Summaryb
Model	R	R Square	Adjusted R Square	Std. Error of the Estimate
1	.248a	.062	.059	1.65621
a. Predictors: (Constant), Hyper
b. Dependent Variable: GADD

ANOVAa
Model	Sum of Squares	df	Mean Square	F	Sig.
1	Regression	62.512	1	62.512	22.789	.000b
Residual	951.826	347	2.743
Total	1014.338	348
a. Dependent Variable: GADD
b. Predictors: (Constant), Hyper

Coefficientsa
Model	Unstandardized Coefficients	Standardized Coefficients	t	Sig.
B	Std. Error	Beta
1	(Constant)	3.337	.197		16.981	.000
Hyper	.062	.013	.248	4.774	.000
a. Dependent Variable: GADD

From the above, the linear regression equation comes out as follows.

GADD = 3.337 + 0.062HYPER

Even though the R² value is quite low but the slope and model seems significant. The R² value is 0.062 which implies that only 6.2% of the variations in the GADD variable are explained by corresponding changes in HYPER variable. Possibly, the results could have improved to some extent if the four outliers would have been excluded. The relationship between the variables is quite weak as the model requires additional of other independent variables to improve the predictive power (Eriksson. & Kovalainen, 2015).

The standardised residuals plot is indicated as follows.

The above can be assumed to be approximately normal especially if the outliers are eliminated from the data.

Further, the regression analysis was run for different years of students to highlight if there was any significant difference in results but it was not so as is apparent from the following output where the coefficient of regression for each of the years remained less than 10%.

YEAR 1

Model Summary
Model	R	R Square	Adjusted R Square	Std. Error of the Estimate
1	.311a	.096	.083	1.56942
a. Predictors: (Constant), Hyper

YEAR 3

Model Summary
Model	R	R Square	Adjusted R Square	Std. Error of the Estimate
1	.131a	.017	.005	1.52697
a. Predictors: (Constant), Hyper

YEAR 5

Model Summary
Model	R	R Square	Adjusted R Square	Std. Error of the Estimate
1	.157a	.025	.008	1.35552
a. Predictors: (Constant), Hyper

The anxiety score has been predicted using the given data and regression models for the given hyperactivity impulse score. The relevant output is shown below.

It is apparent from the above that the point estimate is 4.08 while the LCL and UCL are 0.82074 and 7.34543 respectively. It is apparent that the prediction has not worked considering the actual value does not fall within the interval estimated (Flick, 2015).

Conclusions

Based on the above discussion, it is apparent that the correlation between the hyperactivity impulse score and anxiety score is weak but significant. Similarly, the linear regression model does not form a good fit owing to low predictive power but does indicative that independent variable is significant.  The answer has been obtained to the research question but further research is required to solicit more clarity and thereby enhance the utility of the results.  A key limitation of the given sample was that the control group was smaller than the affected sample and hence the given results may be more indicative of the people suffered from ADHD and may not apply for the general population. This is apparent from the reduced data where the representation of ADHD patients is even more exaggerated.

References

Eriksson, P. & Kovalainen, A. (2015) Quantitative methods in business research. 3rd ed. London: Sage Publications.

Flick, U. (2015) Introducing research methodology: A beginner’s guide to doing a research project. 4th ed. New York: Sage Publications.

Hillier, F. (2016) Introduction to Operations Research. 6th ed. New York: McGraw Hill Publications.

Provatt, F., Dehlil, V., Taylor, N. & Marshall, D. (2015) Anxiety in College Students With ADHD: Relationship to Cognitive Functioning, Journal of Attention Disorders, 19(3) 222 –230