Harmless omission in the standardization of demographic rates.

"Standardization is a well-known technique used to avoid compositional effects when schedules of demographic rates are compared for two or more subpopulations. Common sense tells us that such standardization can be omitted when the subpopulations have the same structure with respect to the covariates one could standardize for. The present paper gives a theoretical justification of this intuitive insight and relates it to the theory for harmless model mis-specification in intensity-regression analysis. The idea of the latter notion is that under certain circumstances one can omit factors without producing biases which affect the coefficients of remaining covariates, even when the omitted factors genuinely affect the investigated behavior." The method is illustrated using risk of divorce. (SUMMARY IN FRE)