We tested the difference in frequency of KIR/HLA haplotypes using the Fisher exact test. To avoid type I error, we calculated an adjusted p value with Bonferroni correction and used the resulting p value (p<0.02) to determine significance. We conducted binary logistic regression to assess the effect of HLA and KIR genotypes on the clinical outcome of patients with EVD. We selected any variable with a significant univariate test at a relaxed p value of 0.25 as a candidate for the multivariate analysis. This criterion enabled us to reduce the initial number of variables (i.e., genes) in the model, while simultaneously reducing the risk of missing important variables (29,30). After preselection, we built a binomial logistic regression model that comprised all remaining explanatory variables and performed backward elimination. The model simultaneously used Bayesian information criterion (31) and Fisher exact tests at the 5% significance level.

As an extension to the binomial analysis, we conducted multinomial logistic regression. This model assessed the effect of HLA and KIR genotypes on the clinical outcome of patients with EVD and determined the KIR gene profile associated with these outcomes. We conducted Bayesian model averaging (BMA) using the R packages BMA and mlogitBMA (32,33). These packages enabled us to account for uncertainty about the explanatory variables using a Bayesian information criterion approximation to the posterior model probabilities. After removing variables that generated collinearity issues, we searched the model space using the fast leaps and bounds algorithm (34). As the first step in applying BMA to solve the variable selection problem for multinomial logit data, mlogitBMA uses the approach of Begg and Gray (35), which approximates large-scale multinomial logistic regressions as a series of binary logistic regressions (36).

Note: The content above has been extracted from a research article, so it may not display correctly.

Please log in to submit your questions online.
Your question will be posted on the Bio-101 website. We will send your questions to the authors of this protocol and Bio-protocol community members who are experienced with this method. you will be informed using the email address associated with your Bio-protocol account.

We use cookies on this site to enhance your user experience. By using our website, you are agreeing to allow the storage of cookies on your computer.