| Draper, D.: Inference and hierarchical modeling in the social sciences. Journal of Educational and Behavioral Statistics 20 (1995) 115--147 |
....the variances are near zero and have highly skewed posterior distributions. In some cases, the usual point estimates (including the ones described in this paper) of the hierarchical variance are zero (see, for example, the educational testing study described in Rubin, 1981, and also discussed by Draper, 1995, and Gelman et al. 1995, x5.5) Setting a variance parameter to zero is generally not desirable because it would lead to falsely precise estimates of random effects. Setting the variance to some nonzero value would require additional work which, in practice, would not be done since it would ....
Draper, D. (1995). Inference and hierarchical modeling in the social sciences (with discussion).
....the (weighted average) placebo mortality of 11.5 (this di erence is highly clinically signi cant) However, the sixth and largest trial, AMIS, went the other way: an increase of 1.2 percentage points in aspirin mortality (a 12 rise from the placebo baseline of 9. 7 ) Some relevant questions (Draper, 1995): Q1 Why did AMIS get such di erent results Q2 What should be done next to reduce the uncertainty about Q 1 Q 3 If you were a doctor treating a patient like those eligible for the trials in Table 2.1, what therapy should you employ while answers to Q1 and Q2 are sought One possible ....
Draper D (1995). Inference and hierarchical modeling in the social sciences (with discussion). Journal of Educational and Behavioral Statistics, 20, 115-147, 233-239.
No context found.
Draper, D.: Inference and hierarchical modeling in the social sciences. Journal of Educational and Behavioral Statistics 20 (1995) 115--147
Online articles have much greater impact More about CiteSeer.IST Add search form to your site Submit documents Feedback
CiteSeer.IST - Copyright Penn State and NEC