K Azam , A Gerami , K Mohammad , A Kazemnejad ,
Volume 2, Issue 1 (4-2004)
Abstract
In large–scale sampling opeartions (e.g. nation-wide health surveys) we always face the problem of non-response item(s) and/or non-response unit(s). In fitting a model to the data we have two groups of variables, namely dependent and independent variables. Non-response may occur for any of these groups of variables. In this paper we assume Y as a categorical dependent variable with three levels, Z and X as independent variables from any kind: scale, categorical, ordinal, etc. We have complete data on the first two variables and we assume that the missing items follow a random pattern (MAR). Then a model is devised based on the likelihood function for the whole data set (including missing values) and estimates of parameters are compared with those obtained by statistical programs such as SPSS, which are only based on completely observed data and ignore units with missing data. Our results show that the likelihood-based model is superior to the standard approach utilized by the software packages. The comparison is made using data on thyriod disease (goiter) obtained by a health survey in Gazvin province.
M Karimlou , K. Mohammad , M. R Meskhani , G.r Jandaghi , K Nouri , E Pasha , K Azam ,
Volume 4, Issue 2 (5-2006)
Abstract
Background and Aim: Logistic regression is an analytic tool widely used in medical and epidemiologic research. In many studies, we face data sets in which some of the data are not recorded. A simple way to deal with such "missing data" is to simply ignore the subjects with missing observations, and perform the analysis on cases for which complete data are available.
Materials and Methods: We consider methods for analyzing logistic regression models with complete data recorded for some covariates (Z) but missing data for other covariates (X). When data on X are Missing At Random (MAR), we present a likelihood approach for the observed data that allows the analysis as if the data were complete.
Results: By this approach, estimation of parameters is done using both Maximum Likelihood and Bayesian methods through the Markov Chain Monte Carlo numerical computation scheme and the results are compared.
The illustrative example considered in this article involves data from lung auscultations as part of a Health Survey in Tehran.
Conclusion: In comparing different methods, Bayesian estimates using the model described in this study are much closer to those generated by analysis of the full data by the standard model.
