Scandinavian Working Papers in Economics
Working Papers in Economics,
University of Gothenburg, Department of Economics
No 788: Serial-correlation-robust power calculation for the analysis-of-covariance estimator
Claes Ek ()
Additional contact information
Claes Ek: Department of Economics, School of Business, Economics and Law, Göteborg University, Postal: P.O. Box 640, SE 40530 GÖTEBORG, Sweden
Abstract: Ex-ante power calculation is an essential part of the toolkit of experimental economics. In panel experiments, analysis of covariance (ANCOVA) is more efficient than differencein- differences and is often preferred. The present paper derives a general serial-correlationrobust variance formula and provides the first analytical ANCOVA power-calculation framework to work with real data. A related earlier procedure for ANCOVA was found by Burlig et al. (2020) to yield incorrect power when used to calibrate a minimum detectable effect on actual data, and the authors caution against using it in practice. I show that these errors arose because time shocks were not properly accounted for in an intermediate procedure estimating residual-based variance parameters from pre-existing data. My procedure resolves such issues, thus providing a framework for accurate power calculation with ANCOVA.
Keywords: power calculation; experimental design; panel data; ANCOVA
38 pages, First version: June 2020. Revised: November 2021. Earlier revisions: October 2020, October 2020, November 2021.
Full text files
65185 HTML file Full text
Questions (including download problems) about the papers in this series should be directed to Ann-Christin Räätäri Nyström ()
Report other problems with accessing this service to Sune Karlsson ().
This page generated on 2024-02-05 17:11:23.