I have a response variable with 2 categories and $500$ predictor variables. The $500$ coefficient $a_1, a_2, a_3, \ldots, a_{500}$ ranges $(-1, 1)$. Positive $a_i$ indicates category A; negative $a_i$ indicates category B. The larger the coefficient, the stronger it indicates its correspondent category. I get the coefficient from a researcher, who score each coefficient from -1 to 1 based on the importance and influence in classification.

To classify an object that have attributes $x_1, x_2, x_3,\ldots,x_{500}$, I am thinking of using logistic regression. But I do not know how to deal with continuous data (the range of the coefficient is continuous from -1 to 1). Is logistic regression viable?

If not, will someone help with other methods and post your code? I prefer using R.

  • $\begingroup$ Are you saying that you have a response variable with 500 categories, or a response variable with 2 categories and 500 predictor variables? Also, if you haven't fit a model yet, where did the coefficients come from? You ask about continuous data, are these response variables or covariates? $\endgroup$ – gung - Reinstate Monica Jul 23 '12 at 15:48
  • $\begingroup$ What is your dependent variable? A number ranging from $(-1,1)$? $\endgroup$ – Macro Jul 23 '12 at 15:48
  • $\begingroup$ Whether logistic regression is viable depends (in large part) on the nature of the dependent variable. It sounds like you have a dichotomous DV, which would be right for logistic regression. There may be other problems - 500 predictors is very rarely going to be sensible. $\endgroup$ – Peter Flom Jul 23 '12 at 15:57
  • $\begingroup$ By "rarely going to be sensible," do you mean the predictors are too many? For each object, about 10 to 20 xi are larger than 0, the others = 0. $\endgroup$ – juju Jul 23 '12 at 16:01
  • $\begingroup$ @Macro I have revised the question. Thanks in advance. $\endgroup$ – juju Jul 23 '12 at 16:09

Browse other questions tagged or ask your own question.