Quasi-Variance
editQuasi-variance (qv) estimates are a statistical approach to overcome the reference category problem when estimating the effects of a categorical explanatory variable within a statistical model.
Quasi-variances are approximations of variances. Quasi-variances are statistics associated with the parameter estimates (coefficients) of the different levels of categorical explanatory variables within regression models. Quasi-variances should routinely be presented alongside parameter estimates to enable readers to assess differences between any combinations of parameter estimates for a categorical explanatory variable. The approach is beneficial because such comparisons are not usually possible without access to the full variance-covariance matrix for the estimates.
The underlying idea was first proposed by Ridout[1] but the technique was set out by Professor David Firth[2].
The suitability of this technique for social science data analysis has been demonstrated[3].
An on-line tool for the calculation of quasi-variance estimates is available and a short technical description of the methodology is provided.
Quasi-variances can be calculated in Stata using the QV module[4] and can also be calculated in R using the package qvcalc. An extended set of resources are with examples in Stata and SPSS are also available.
References
edit- ^ Ridout, M.S. (1989). Summarizing the Results of Fitting Generalized Linear Models to Data from Designed Experiments. New York: Springer-Verlag. pp. 262—9.
- ^ Firth, David (2016-06-24). "1. Overcoming the Reference Category Problem in the Presentation of Statistical Models". Sociological Methodology. 33 (1): 1–18. doi:10.1111/j.0081-1750.2003.t01-1-00125.x.
- ^ Gayle, Vernon; Lambert, Paul S. (2007-12-01). "Using Quasi-variance to Communicate Sociological Results from Statistical Models". Sociology. 41 (6): 1191–1208. doi:10.1177/0038038507084830. ISSN 0038-0385.
- ^ Chen, Aspen (2014-07-21), QV: Stata module to compute quasi-variances, retrieved 2017-03-15