I am doing logistic regression with 3 attributes. According to my data set I am expecting all coefficients to be positive. But it gives me both positive and negative coefficients. Is it possible to have all positive coefficients using logistic regression.
4
-
$\begingroup$ Are you expecting the coefficients on the log odds scale to be positive? $\endgroup$– Demetri PananosCommented Feb 7, 2019 at 5:56
-
$\begingroup$ no, i want the logistic regression coefficients to be positive $\endgroup$– jenyKCommented Feb 7, 2019 at 6:17
-
$\begingroup$ Do you have any regression software that allows setting bounds on the parameters? If so, you could set all of the lower bounds at zero. $\endgroup$– James PhillipsCommented Feb 7, 2019 at 12:25
-
$\begingroup$ Can you post the output of your software? $\endgroup$– Demetri PananosCommented Feb 7, 2019 at 13:34
Add a comment
|
1 Answer
$\begingroup$
$\endgroup$
Yes, it is possible to have all positive coefficients in logistic regression.
That doesn't say very much, if a covariable $x$ have a negative coefficient you can just replace that covariable by $-x$ and get a positive coefficient ... you would get a more useful answer if you told us about your real problem, your data, the context and research questions.
answered Mar 28, 2019 at 22:17