0
$\begingroup$

Say you have data that looks like this:

enter image description here

How would you predict y? I think y looks like some sort of (iid) exponential distribution function, but I can't really find any way to model this in R. Thanks in advanced for any suggestions!

$\endgroup$
  • 1
    $\begingroup$ It's no use looking at the marginal distribution when the model is for the conditional distribution. About the only time looking at the histogram of the response works okay is when the correct model for the mean is constant (& you have more bins in the plot). An exponential might work but you can't tell from this plot. If you think the conditional distribution should be exponential you could try fitting a glm and checking residuals, but it depends on what y&x are. If you don't have a suitable distributional model, consider pulling off a randomly selected subset to explore I'd start with logs $\endgroup$ – Glen_b Feb 28 '17 at 16:48
0
$\begingroup$

If your goal is to predict y I would try to do a few transformations to see if I can make the data linear. Try log, sqrt and the inverse transformation to start with.

$\endgroup$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.