Gerhard Osius | Searcharxiv

arXiv2012

The Asymptotic Covariance Matrix of the Odds Ratio Parameter Estimator in Semiparametric Log-bilinear Odds Ratio Models

The association between two random variables is often of primary interest in statistical research. In this paper semiparametric models for the association between random vectors X and Y are considered which leave the marginal distributions arbitrary. Given that the odds ratio function comprises the whole information about the association the focus is on bilinear log-odds ratio models and in particular on the odds ratio parameter vector θ. The covariance structure of the maximum likelihood estimator θ^ of θ is of major importance for asymptotic inference. To this end different representations of the estimated covariance matrix are derived for conditional and unconditional sampling schemes and different asymptotic approaches depending on whether X and/or Y has finite or arbitrary support. The main result is the invariance of the estimated asymptotic covariance matrix of θ^ with respect to all above approaches. As applications we compute the asymptotic power for tests of linear hypotheses about θ - with emphasis to logistic and linear regression models - which allows to determine the necessary sample size to achieve a wanted power.

math.ST↗

arXiv2009

Asymptotic inference for semiparametric association models

Association models for a pair of random elements $X$ and $Y$ (e.g., vectors) are considered which specify the odds ratio function up to an unknown parameter $\boldsθ$. These models are shown to be semiparametric in the sense that they do not restrict the marginal distributions of $X$ and $Y$. Inference for the odds ratio parameter $\boldsθ$ may be obtained from sampling either $Y$ conditionally on $X$ or vice versa. Generalizing results from Prentice and Pyke, Weinberg and Wacholder and Scott and Wild, we show that asymptotic inference for $\boldsθ$ under sampling conditional on $Y$ is the same as if sampling had been conditional on $X$. Common regression models, for example, generalized linear models with canonical link or multivariate linear, respectively, logistic models, are association models where the regression parameter $\boldsβ$ is closely related to the odds ratio parameter $\boldsθ$. Hence inference for $\boldsβ$ may be drawn from samples conditional on $Y$ using an association model.

math.ST↗