0

Say we have a model with three predictors and one dependent variable. For the sake of our example let the model be the following: Y = X1 + X2 + X3 + e

Where Y is a 1-10 feeling towards apples, and X1, X2, and X3 is information provided to respondents about apples. The purpose of this model is to estimate how different types of information affects an individual's feelings towards apples.

I am wondering if it is possible to measure the effect of each predictor on the variance of the outcome, i.e. feelings towards apples. Say I want to know whether X1 increases or decreases the variance of Y. It seems difficult, because var(Y) is constant throughout Y.

Note: Say X1, X2, and X3 are randomized (if this might change anything).

juanjedi
  • 1
  • 1
    Do you have data or is this a purely theoretical question? Sample size? (you will need a larger sample to model the variance than for only modeling the conditional expectation of Y). Show us some residual plots? Some similar posts: https://stats.stackexchange.com/questions/273516/anova-for-5-likert-scale-comparing-3-4-groups, https://stats.stackexchange.com/questions/389127/does-it-make-sense-to-use-residuals-as-an-independent-variable/390592#390592, – kjetil b halvorsen Sep 07 '22 at 18:40
  • I do not have the data, yet. But yes, this would be a practical question for a future application with data. – juanjedi Sep 07 '22 at 18:49
  • You could try to model both, conditional mean and conditional variance as functions of the predictors or use quantile regression. – utobi Sep 07 '22 at 20:00

0 Answers0