Accounting for non-normal covariates in treatment effects from count regressions

Principal Investigator(s): View help for Principal Investigator(s) Christoph Kiefer, Bielefeld University, Germany

Name	File Type	Size	Last Modified
dACTIVE.rds	application/gzip	15 KB	12/15/2020 02:12:AM
data_import.R	text/x-r-syntax	2.7 KB	12/15/2020 02:11:AM
supplementA.pdf	application/pdf	329.4 KB	02/02/2021 09:07:PM

Project Citation:

Kiefer, Christoph. Accounting for non-normal covariates in treatment effects from count regressions. Ann Arbor, MI: Inter-university Consortium for Political and Social Research [distributor], 2021-02-16. https://doi.org/10.3886/E128941V2

Project Description

Summary:

This project contains supplementary material for Kiefer & Mayer (2021).

Abstract of the manuscript:

The effects of a treatment or an intervention on a count outcome is often of interest in applied research. When controlling for additional covariates, a negative binomial regression model is usually applied to estimate conditional expectations of the count outcome. The difference in conditional expectations under treatment and under control is then defined as the (conditional) treatment effect. While traditionally aggregates of these conditional treatment effects (e.g., average treatment effects) are computed by averaging over the empirical distribution, a recently proposed moment-based approach allows for computing aggregate effects as a function of distribution parameters. The moment-based approach makes it possible to control for (latent) multivariate normally distributed covariates and provides more reliable inferences under certain conditions.

In this project, we propose three different ways to account for non-normally distributed continuous covariates in this approach: (a) an alternative, known non-normal distribution, (b) a plausible factorization of the joint distribution, and (c) an approximation using finite Gaussian mixtures. A saturated model is used for categorical covariates, making a distributional assumption obsolete.

For illustration of our methods, we used a subset of the ACTIVE dataset:

Tennstedt, Sharon, Morris, John, Unverzagt, Frederick, Rebok, George W., Willis, Sherry L., Ball, Karlene, and Marsiske, Michael. Advanced Cognitive Training for Independent and Vital Elderly (ACTIVE), United States, 1999-2001. Inter-university Consortium for Political and Social Research [distributor], 2010-06-30. https://doi.org/10.3886/ICPSR04248.v3

Citation of project publication:

Kiefer, C. & Mayer, A. (2021). Treatment effects on count outcomes with non-normal covariates. British Journal of Mathematical and Statistical Psychology. doi: 10.1111/bmsp.12237

Along with this project comes our R package CountEffects, which can perform effect analysis according to the three proposed approaches on the dACTIVE.rds file provided in this project. The Supplemental Material to our paper describes how CountEffects can be installed and used.

Scope of Project

Subject Terms: treatment effects; non-normal covariates; negative binomial regression

Methodology

Data Source:

ACTIVE dataset at ICPSR:

additional dataset for baseline items of CES-D via personal communication

Related Publications

Download this project

Published Versions

V2 [2021-02-16]

V1 [2020-12-16]

Export Metadata

Dublin Core

DDI 2.5

Report a Problem

Found a serious problem with the data, such as disclosure risk or copyrighted content? Let us know.

This material is distributed exactly as it arrived from the data depositor. ICPSR has not checked or processed this material. Users should consult the investigator(s) if further information is desired.