Logit effect values and efficiency of significance tests when a noncollapsible variable is omitted

For linear models, randomization of the assignment to the levels of a variable A is a sufficient condition to obtain correct estimates of the effect of A on another variable B. For logit models, the omission of a relevant variable Z often results in invalid estimates of the effect of A and in a loss of power of the significance test. The values of this underestimation and of the loss of power are evaluated here asymptotically and from simulation studies. Consequences for practical use of logit models in randomized settings are underlined.     

A numerical and experimental study of compliance and collapsibility of preterm lamb tracheae

Knowledge of the mechanical behaviour of immature tracheae is crucial in order to understand the effects exerted on central airways by ventilatory treatments, particularly of Total Liquid Ventilation. In this study, a combined experimental and computational approach was adopted to investigate the compliance and particularly collapsibility of preterm lamb tracheae in the range of pressure likely applied during Total Liquid Ventilation (−30 to 30 cmH₂O). Tracheal samples of preterm lambs (n=5; gestational age 120–130 days) were tested by altering transmural pressure from −30 to 30 cmH₂O. Inflation (S_i) and collapsing (S_c) compliance values were calculated in the ranges 0 to 10 cmH₂O and –10 to 0 cmH₂O, respectively. During the tests, an asymmetric behaviour of the ΔV/V₀ vs. P curves at positive and negative pressure was observed, with mean S_i=0.013 cmH₂O⁻¹ and S_c=0.053 cmH₂O⁻¹. A different deformed configuration of the sample regions was observed, depending on the posterior shape of cartilaginous ring. A three-dimensional finite-element structural model of a single tracheal ring, based on histology measurements of the tested samples was developed. The model was parameterised in order to represent rings belonging to three different tracheal regions (craniad, median, caudal) and numerical analyses replicating the collapse test conditions were performed to evaluate the ring collapsibility at pressures between 0 and −30 cmH₂O. Simulation results were compared to experimental data to verify the model's reliability. The best model predictions occurred at pressures −30 to −10 cmH₂O. In this range, a model composed of median rings best interpreted the experimental data, with a maximum error of 2.7%; a model composed of an equal combination of all rings yielded an error of 12.6%.     

Inference on Collapsibility in Generalized Linear Models

GREENLAND and MICKEY (1988) derived a closed-form collapsibility test and confidence interval for IxJxK contingency tables with qualitative factors, and presented a small simulation study of its performance. We show how their method can be extended to regression models linear in the natural parameter of a one-parameter exponential family, in which the parameter of interest is the difference of “crude” and “adjusted” regression coefficients. A simplification of the method yields a generalization of the test for omitted covariates given by HAUSMAN (1978) for ordinary linear regression. We present an application to a study of coffee use and myocardial infarction, and a simulation study which indicates that the simplified test performs adequately in typical epidemiologic settings.     

Design Aspects and Odds Ratio Estimation in Multicenter Case-Control Studies

In this paper we propose a method to be used in the planning stage of a case-control study. An allocation rule for controls in multicenter case-control studies is proposed which would assure a simple, efficient and unbiased estimation of the odds ratio in the pooled data. It is shown that the efficiency of the design increases with increasing correlation between study center and risk factor. Sources of bias and their implications for relative risk estimation are discussed. The method is demonstrated with data from a case-control study.