1
$\begingroup$

I am running a logistic regression model with the following data and variables:

Independent variables (characteristics of a person coded as dummy)

  • Test score (continuous variable between 0 and 1)
  • Education (dummy variable for Ivy league degree; categorical variable for type of major)
  • Age (discrete variable)

Dependent variable (person's performance at task)

  • Task is to select a company that will be successful at given point in time (success is coded as dummy variable: 0: fail, 1: success -- based on a definition from the specific literature)

I want to run the regression such that each combination of person-task (task meaning company selected) is one observation.

There are 2 issues I am not sure how to account for in the model:

  1. Most persons completed a different amount of tasks at different points in time (some selected only 5 companies, others 60), thus I am concerned that those with more observations will be weighted more heavily.

  2. Some persons selected the same companies, thus they share the same outcome (success or failure) and hence, I am concerned that some outcomes are overweighted if a company was chosen several times by different persons (one person cannot choose the same company more than once).

How can I account for this in the logistic regression model? I have been looking at using weights and clustering standard errors, but I am not sure if this is entirely correct.

Thank you in advance for your help!

$\endgroup$
  • $\begingroup$ Can you explain the situation more? What exactly is the task? What are they selecting a company for? Is your outcome whether they were right in their selection? What does that mean? Most likely you should be doing something different, but I can't tell what. $\endgroup$ – gung Sep 7 '18 at 19:09

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Browse other questions tagged or ask your own question.