I have two independent continuous variables like Age, Price and an outcome variable like purchased or not purchased
Now, which test should I use to ascertain the association of continuous input variable with categorical outcome variable?
$\begingroup$
$\endgroup$
5
-
$\begingroup$ isn't like a classic differential test ? if you have normalized distributed feature one can use t-test, otherwise sth like wilcoxon - the result will be whether the continues value is enriched in one group that the other ?! $\endgroup$– user702846May 25, 2021 at 20:30
-
$\begingroup$ Can't we use point biseral corelation? $\endgroup$– The GreatMay 26, 2021 at 0:07
-
$\begingroup$ your question is more statistical - I recommend to consult in stats.stackexchange.com - meanwhile : correlation coefficients or coefficients from logistic regression only give you the "effect size" not the level of significance. you have used term : "association" which implies you are after finding the level of significance :-) in stat words matter. $\endgroup$– user702846May 26, 2021 at 7:38
-
$\begingroup$ @user702846 - let's I have one binary variable and continuous variable but the values for continuous variable are only very few. Meaning, ex: in the whole column, only 5 values keep repeating for continuous variable. In this case, can I consider them as categorical (because oy 5 unique values/levels) and apply chi-square test for association? $\endgroup$– The GreatJul 4, 2021 at 3:28
-
$\begingroup$ technically you can - but what it means and how to interpret I don't know. $\endgroup$– user702846Jul 4, 2021 at 20:46
Add a comment
|
1 Answer
$\begingroup$
$\endgroup$
3
Try logistic regression. The sigmoid function and its variants allow for conversion of continuous inputs into a binary outcome variable.
-
$\begingroup$ Can log reg work fpr just 1 input variable and 1 outpit variable? $\endgroup$ May 25, 2021 at 23:56
-
$\begingroup$ cant we use point biseral corelation? Upvoted for thehelp $\endgroup$ May 26, 2021 at 0:08
-
$\begingroup$ Perhaps yes, but also try out f-classif, mutual-info-classif under sklearn.feature_selection $\endgroup$ May 26, 2021 at 6:45