Tuesday, November 17, 2009

Revise and Resubmit...

I got another revise and resubmit today. This time for my paper on interfuel elasticities of substitution. The most important point the referees want me to address is the weights I use in the meta-regression. I use the square root of sample size. They would like me to use the standard errors of the parameters from the original studies. However, one referee admits that when the statistics that are being compared in a meta-analysis are complicated nonlinear functions of the original estimated parameters as is the case here this may in any case not be desirable and provides a reference for backing up my view.

For example, some authors only provide standard errors for the original parameters that they estimate, not the elasticities they present, which are already nonlinear functions of the original parameters. In order to derive an estimate of the standard error of the elasticity I need to know the covariance between the parameters (which is almost never presented) or assume it is zero, which introduces another source of error into my study. In any case, my study uses shadow elasticities of substitution, which are share weighted means of the Morishima elasticities of substitution. In many or most cases, authors do not present the cost shares and I calculate them from the authors' parameter estimates. So even my cost shares are non-linear functions of the original parameters.

No comments:

Post a Comment