Choices other than OLS for Impact Estimation

This forum is closed, and read-only.
Locked
Joynerjd
PLS Junior User
Posts: 3
Joined: Mon Mar 18, 2013 12:41 pm
Real name and title:

Choices other than OLS for Impact Estimation

Post by Joynerjd »

It would be helpful in some applications to have a choice in the method used to estimate impacts between LVs.

For instance, a robust regression estimator that down-weighted outliers could be helpful, as might a PLS regression estimator that helped to overcome high collinearity among predictor LVs. And a quantile regression estimator that allowed one to predict the conditional median (or other quantile) rather than the conditional mean could also be helpful.

While the choice could be set up to be global (across all inner equations), there might be instances where a local choice (specific to a particular inner equation) would be very helpful and provide greater flexibility.

Moving in this direction, you could also potentially include a logistic regression estimator for use on terminal dependent LVs (who are not also predictors elsewhere in the model) that would further expand the functionality of the platform and allow the "proper" use of binary outcomes.
Jason Joyner, Ph.D.
Business Analysis
Chick-fil-A, Inc.
USA
Locked