1
$\begingroup$

This is covariate age in my logistic regression. How should I treat it? Gets a little insecure. Have tried to read, but still insecure. Right now I treat it as if it were linear. A polynomial is not appropriate. Some tips? I have a binary response variable

enter image description here

x axis is age, the y axis is the percentage of approved students first semester

$\endgroup$
2
  • 1
    $\begingroup$ "in R" appears irrelevant here. It's common in many fields to use age and age-squared as predictors. The point is not whether you expect a turning point but rather that the quadratic allows some curvature. If the squared term doesn't add much predictability, omit it. $\endgroup$
    – Nick Cox
    Apr 23, 2015 at 9:30
  • 1
    $\begingroup$ Some other possibilities -- you could start with a linear term (on the scale of the linear predictor, presumably logit scale) and see if that's adequate, or you could fit some GAM-style term, such as a spline. Which of the various suggestions you might pick may depend on what you know or what you want to know, or what you're prepared to assume. $\endgroup$
    – Glen_b
    Apr 23, 2015 at 9:52

0

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Browse other questions tagged or ask your own question.