Dear scholars, I am using the SDQ-25 questionnaire to measure emotional and behavioral problems. I plan to use the ordered logit model for analyses but it violates the rule. is it possible to dichotomize and use binary logistic regression?
The short answer is no. It is best to know what you are doing. At least google the methods and see what they're about and go from there. I attached some excellent course notes. Look up scales of measurement. Interval is not categorical. Good luck, David Booth
SDQ -25 has 4 subscales: Emotional. Conduct, Hyperactivity, and Peer problems. also, a total score can be obtained. There are Parent, teacher, and Self-administered versions. Each of the subscales is rated on a likert scale. and the total of each subscale is calculated. Each subscale represents a separate dimension and there is no ordering i.e. it is not an ordinal scale.
As you get a unique score for each individual on each of the dimensions, it is an interval scale measure. Linear regression analysis is possible on total scores or separately for each dimension.
It is better not to suitably use a scale than what is given in the manual for the intended purpose.
I thank Isam, David and Anand for your responses to the question brought by Gari. I agree with the comments of the above-mentioned scholars. Really, SDG-25 is a psychometric instruments that is scored using a Likert Scale. This tool mostly used among individuals aged 2 to 18 years has 4 sub-scales namely emotional, conduct, hyperactivity and peer problems. It, of course, consists of continuous or numerical values. But some authors established that it may have 5 scales that are emotional symptoms, conduct problems, hyperactivity-inattention, peer problems, and prosocial behaviors. Each scale has 5 items. This means it does not have categories and no study indicated that it may be categorical. Although some prior studies stated that the cut-off (>90th percentile) indicate that the individuals are considered abnormal, the current SDG-25 may not be categorical variable. It may better to analyse using the simple linear regression, multivariate linear regression models, normality test, factorial analysis, or goodness of fit model, and other advanced test. In this case, you will need to make your independent variables continuous for easing the computation of the statistical analysis. Please, try to indicate its psychometric properties as well. But remember that the SDG-25 may be used according to the age category!
For further information you may read the following document: https://doi.org/10.1186/s12955-018-0948-1