I am developing a factor model to predict price change. Bayesian Logistic Regression seems more advanced than the traditional Logistic Regression. However, as in this matlab example
http://cn.mathworks.com/help/stats/exam ... model.html
Each feature vector(the weight of car in the above example) has several tests to evaluate the binomial probability (a number of poor cars in the above example). But in the financial world, to forecast asset A up or down, each feature vector (say, the S&P index, rate, etc.) can only be mapped to only one label by using historical data. In the latter case, I find Bayesian Logistic Regression gives a high bias in parameter estimations. So can can anyone share some experience of using Bayesian Logistic Regression in financial world?