Biased Estimates of Treatment Effect in Randomized Experiments with Nonlinear Regressions and Omitted Covariates

Abstract

JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide range of content in a trusted digital archive. We use information technology and tools to increase productivity and facilitate new forms of scholarship. For more information about JSTOR, please contact support@jstor.org.. Biometrika Trust is collaborating with JSTOR to digitize, preserve and extend access to Biometrika. SUMMARY Certain important nonlinear regression models lead to biased estimates of treatment effect, even in randomized experiments, if needed covariates are omitted. The asymptotic bias is determined both for estimates based on the method of moments and for maximum likelihood estimates. The asymptotic bias from omitting covariates is shown to be zero if the regression of the response variable on treatment and covariates is linearor exponential, and, in regular cases, this is a necessary condition for zero bias. Many commonly used models do have such exponential regressions; thus randomization ensures unbiased treatment estimates in a large number of important nonlinear models. For moderately censored exponential survival data, analysis with the exponential survival model yields less biased estimates of treatment effect than analysis with the proportional hazards model of Cox, if needed covariates are omitted. Simulations confirm that calculations of asymptotic bias are in excellent agreement with the bias observed in experiments of modest size. Some key words: Bias; Clinical trial; Nonlinear model; Omitted covariate; Randomization. J O E{H(y I T, X) exp (oc*T)} d where E is the usual expectation operator over T, X and Y. In principle, (4-15) may be solved for oc* for each fixed (oc, /B) to obtain the exact bias. We illustrate this calculation for exponential data with uniform entry on [0, Y0] and analysis at Y0. Without loss of generality, AO(w) = I because, from (41), A(w I X, T) = A0 exp (j) exp (ocT + ?X). For fixed T, X, H(yI|T, X) = pr [Y > yI|T, X] = exp {-exp (C~)y} (1-y/Y0). This content downloaded from 91.229.248.104 on Wed

Binary Regression in Truncated Samples, with Application to Comparing Dietary Instruments in a Large Prospective Study

“…If σ 2 R is the variance of C given ( F , X ), then by the usual probit approximation to the logistic, When β 3,trun σ R is not large, the denominator inside is close to 1.0, see Carroll et al (2006, Chapter 4.8). Gail et al (1984) give a more general calculation when R is independent of ( F , X ), with a similar conclusion. Gail et al also show that if regression function H (·) is an exponential rather than logistic function, then, because pr( Y = 1 | F , X ) = exp(β 0,trun + F β 1,trun + X T β 2,trun ) × E {exp(β 3,trun R ) | F , X }, estimates based on are consistent if R is independent of ( F , X ), or, more generally, if E {exp(β 3,trun R ) | F , X } is a constant.…”

“…The method uses as its motivation work by Gail, Wieand, and Piantadosi (1984), and requires little more than ordinary linear logistic regression. It posits a risk model in which the truncation variable is included, and then employs a new residualization method to approximate the marginal model when the truncation variable is not included.…”

The performance of different propensity score methods for estimating marginal hazard ratios

“…A measure of effect is collapsible if, in the absence of confounding, the conditional effect and the marginal effect coincide. Differences in means and risk differences are collapsible, whereas odds ratios and hazard ratios are non‐collapsible . Neuhaus et al .…”

“…Linear treatment effects (differences in means and differences in proportions) are collapsible: the conditional and marginal treatment effects will coincide. However, when outcomes are binary or time to event in nature, the odds ratio and the hazard ratio are not collapsible . Rosenbaum has noted that propensity score methods allow one to estimate marginal, rather than conditional, effects .…”

Sensitivity Analysis in Non-Inferiority Trials with Residual Inconstancy After Covariate Adjustment

“…For rare events, however, it would be more natural to consider the relative risk or the odds ratio as the effect measure. It will be of interest to extend the proposed approach to those effect measures with appropriate adjustment for possible non‐collapsibility (Gail et al ., ; Greenland et al ., ).…”