I am doing logistic regression analysis using multiple predictors for a binary outcome.I had about 10 predictors and tried to find the best model using 'glmulti' package in R. I have got a significant model with four predictors. Now I want to explore the effect of each predictor on the outcome according to the model. I thought of keeping the continuous variables at mean values and predicting probabilities for categorical predictors. Also I am trying to compute predicted probabilities for a range of continuous variable at each level of categorical predictor and plot the graph. But I need to see published medical articles where people really explored the multivariate logistic model and explained in detail about the derived model in terms of the contribution of each predictor to the outcome. Can anyone give statistical advice and also refer few articles ?
"Multivariate" means more than one dependent variable. Did you mean multivariable? What do you mean "have got a significant model"? Were all the variables in the model pre-specified? More to your question, see for example http://biostat.mc.vanderbilt.edu/rms; the basic idea is to vary one predictor at a time, hold other predictors constant, and plot the predicted log odds of $Y=1$ vs. the varied predictor. When doing so, it does not matter what values you adjust non-plotted predictors to. I use the median of continuous variables and mode of categorical ones. When transforming the predictions to be on the probability scale, it matters more what are the predictor settings for the non-varying predictor. You can still use the median/mode but it may be worth plotting for other settings.