What are the advantages and disadvantages of using logistic regression to represent the probability of a binary dependent variable Y, given that the independent variables x1, x2, ..., xn are also binary?
5
-
1$\begingroup$ Advantages and disadvantages as compared to what exactly? Compared to other model types? Or compared to using only continuous variables, or a mix of binary predictors and continuous ones, or >2 level categorical predictors? Or are you asking about representing a probability instead of classification? Please elaborate if you think your question is different than the one @kjetilbhalvorsen has referenced to. $\endgroup$– IWSCommented Jun 2, 2017 at 14:57
-
$\begingroup$ The other question asked 'if it is feasible to perform logistic regression with binary dependent and independent variables', which was answered as 'yes'. I guess I am wondering what are the challenges of only using binary independent variables? Does anything change to the logistic regression procedure for only binary independent variables? I am only finding resources where the independent variables are continuous values. Is there a particular name for this type of logistic regression? $\endgroup$– C.NealCommented Jun 2, 2017 at 15:12
-
$\begingroup$ The question remains unclear. What kind of potential "challenges" do you have in mind? Basically, nothing changes between doing LR w/ continuous Xs vs w/ categorical Xs. $\endgroup$– gung - Reinstate MonicaCommented Jun 2, 2017 at 15:19
-
1$\begingroup$ I see, that answers my question, since it is the same procedure. $\endgroup$– C.NealCommented Jun 2, 2017 at 15:29
-
$\begingroup$ If anything changes, then it is in the other direction: using logistic regression with binary covariables have very weak assumptions, linearity is vacuous, but with continuous predictors linearity becomes a strong assumption. $\endgroup$– kjetil b halvorsen ♦Commented Jun 2, 2017 at 17:34
Add a comment
|