Question

Suppose you are a statistician consulting with a researcher who is trying to estimate the effect of a new behavioral health intervention for children presenting with conduct disorders. He has data from two randomized controlled trials, each of which compares the change in behavioral health symptoms between children who received the intervention to those who did not. However, one study measured behavioral health using the Child Behavioral Checklist (CBCL) and the other measured behavioral health using the Eyberg Child Behavior Inventory (ECBI). Although the CBCL and the ECBI are both well-established parent-rated survey measures of child behavioral health symptoms, they have different questions and response options and are scored and interpreted differently. Therefore, a CBCL score and an ECBI score are not meaningfully comparable. The researcher wants to use all of the data he has from both studies to estimate the effect of the intervention.

In a short paragraph (~4-5 sentences), explain what you would advise the researcher to do and why.

Answer 1

Here based on the samples, the person should conduct a hypothesis that the data come from the same population. One way to do it is assume an underlying distribution and test the equality of parameters in the two cases. So if we consider the data to be coming from a particular distribution, we shall conduct test of equality of means and variances. We should scale both the data to the same level and conduct this test.Thus we shall use this scaled data and consider this as the test set and calculate effect of intervention.