Reading: MacCallum, R. C., Zhang, S., Preacher, K. J., & Rucker, D. D. (2002). On the practice of dichotomization of quantitative variables. Psychological methods, 7(1), 19.

Note: I remind you of the statement in the syllabus and that we reviewed on day 1 regarding inclusivity and psychological research. This article makes reference to gender as a dichotomous variable. This is an unfortunate situation, and it is actually ironic, because the authors are advocating not to dichotomize that which is not a true dichotomy. Although not required reading today, here’s a reference that discusses assessment of gender and respecting gender diversity in psychological research: Cameron & Stinson (2019). We will consult this linked source in more detail in the laboratory portion of this course.

1. Paragraph 1-2 is a very quick overview of a few fundamental concepts in psychological research. Key terms mentioned here include: independent variable, dependent variable, correlation, linear regression, and then an extension as described in the paragraph: multiple regression; median, and by extension a median split; rejecting and failing to reject a null hypothesis, normal distribution). Check your understanding by reading this paragraph and seeing how you follow along. If anything in here doesn’t make sense, you will want to review these concepts to understand them.

2. What are ‘quantitative variables’?

   1. Quantitative Variables: any variables where the data represent amounts

3. What is ‘dichotomization’?

   1. Dichotomization: an item or score that initially had a set of continuous values but was then separated into two possible values

   2. Discuss the key issues/problems associated with dichotomization?

      1. Dichotomization used to represent individual differences:

         1. Dichotomization limits the ability to interpret differences within the data and may subjectively group results invalidly.

   3. Statistical issues:

      1. an dichotomization increase the obtained correlation? Decrease it?...or both? Under what circumstances is the obtained correlation more likely to increase following dichotomization?

         1. It could both increase and decrease correlation.

         2. It is more likely to increase following dichotomization if the results are broadly put together Note: understanding 3.b.i. will be easier if you are able to distinguish between a population versus sample correlation (see pp. 25-26).

      2. Effect size --- generally, what happens to effect size following dichotomization?

         1. Effect size decreases?

      3. Power --- generally, what happens to power following dichotomization?

         1. Power increases in the groups but decreases in correlation

4. How common is dichotomization in practice? (at least at the time this article was published? NOTE: Anecdotally, I still see examples of this quite often).

   1. Very common

5. How do researchers justify the practice of dichotomization? (and how do the authors respond to each justification?)

   1. Researchers justify it by saying it was used before, it simplified analysis, gained capability of examining moderator effects, categorized because of skewed data, using significant cutpoints, and improving statistical power.

6. Under what circumstances might it be appropriate to dichotomize?

   1. When using broad age ranges, or other categorical groups that are similar for the content being tested.