Hi, i have a few question. Right now i want to look the interaction between my response variable (Y-axis) which is number of rainy days and my only 1 predictor variable (X-axis) which is amount of rain. Thus, i would like to use sm.regression to look the interaction. So, my question is i really dont understand much about sm.regression, as i read they need bandwidth. is this bandwidth necessary to look out as we our self need to estimate the bandwidth
Is there a particular reason you want to use nonparametric regression rather than standard OLS regression?
Also, while I'm not familiar with this particular method, I'm not sure it does need a bandwidth specified; the examples on the documentation page for the sm.regression function include many in which the function is given only a response and predictor variable.
Also, your question is a little confusing because you say you want to look at the "interaction" between your response and predictor variables. Do you mean "relationship", "association", etc? An "interaction" is between two (or more) predictor variables.
i want i mean is that to look the "relationship" and i want to use non parametric regression as my predictor variables is in continuous data. So, do you think i can just use standard OLS regression?
Yes, having a continuous predictor is not a problem, that's what OLS regression is for.
You can still use a non-parametric regression, but you should have a reason why you're choosing to do it. There are plenty of valid reasons one might use a non-parametric test; but "my predictor is continuous" is not one of them.
if both of my predictor and variable where continuous data, so i still can use OLS regression?
Yes, that is the most common case that OLS regression is used for.
i still had a question sir. for my data, its not normal distribution, so is this OLS regression still can be used?
- Badges
- Science topic