Browsing Faculty of Social Sciences by Subject "Logit model"
Now showing items 1-2 of 2
-
Mokgatlhe, L.; Groenewald, P.C.N. (Elsevier, http://www.elsevier.com, NaN, 2005)[more][less]
Abstract: A method for the simulation of samples from the exact posterior distributions of the parameters in logistic regression is proposed. It is based on the principle of data augmentation and a latent variable is introduced, similar to the approach of Albert and chib (J. Am. Stat. Assoc. 88 (1993) 669), who applied it to the probit model. In general, the full conditional distributions are intractable, but with the introductions of the latent variable all conditional distributions are uniform, and the Gibbs sampler is easily applicable. Marginal likelihoods for model selection can be obtained at the expense of additional Gibbs cycles. The technique is extended and can be applied with nominal or ordinal polychotomous data. URI: http://hdl.handle.net/10311/1127 Files in this item: 1
Mokgatlhe_CSDA_2005.pdf (564.2Kb) -
Amey, KA.A.; Forcheh, N.; Setlhare, K. (Dove Medical Press Ltd. www.dovepress.com/, NaN, 2012)[more][less]
Abstract: Background: Predictive models for mortality due to human immunodeficiency virus (HIV) disease as a result of opportunistic infections, such as tuberculosis and pneumonia, have been developed. Methods: The data are taken from the Statistics South Africa multiple causes of death data for 2006 and 2007, which is available for public use. The dataset was compiled from death notifications, and contains up to five causes of death as well as some demographic characteristics of the deceased. The logistic regression modeling framework was used to model the presence or absence of HIV disease, given the predictive variables. Results: The higher the number of causes listed, the higher the likelihood that HIV would be a cause, with the percentage of notifications of HIV listed increasing from under 2% when only one cause is listed to almost 15% when 4–5 causes are listed. When the logit model was fitted to the multiple cause of death model, it was found that individual demographics were good predictors of the likelihood that the death notification would have HIV as one of the causes of death. Although there are highly significant differences in the likelihood that people of different demographics would die from HIV, the predictive power of these demographic factors on their own is very low, especially when there is only a single cause of death mentioned. With the full multiple cause of death model, two-way interactions between tuberculosis, pneumonia, and other opportunistic infections were highly significant, and their inclusion lead to significant improvements in the predictive power of the model. URI: http://hdl.handle.net/10311/1071 Files in this item: 1
Forcheh_OAMS_2012.pdf (302.7Kb)
Now showing items 1-2 of 2