topic Re: I am wondering if you can help with non-parametric linear regression. in Statistical Procedures

I am wondering if you can help with non-parametric linear regression.

DJGS — Sat, 11 Sep 2021 17:59:51 GMT

I have a set of tooth wear data (from a non-human species) in the form of ratios--- the ratios are the amount of dentine exposure divided by the areas of the chewing surface of the tooth. These ratios are not normally distributed.

I am interested in comparing males and females in terms of their wear ratios as a function of age. So, I have plotted these ratios against the ages of the individuals, and males seem to have a steeper linear slope than do females-- that is, the males wear ratios appear to increase faster as a function of age than do the females' wear ratios. I would like to know the best way to statistically compare males and females in terms of their linear regression slopes but the data, as I say, are not normal. Googling a bit, it seemed that Theil regression might work, but I am not sure. What would you recommend and if you have a recommendation can you point me in the direction of some SAS code that would help? Many thanks!

Re: I am wondering if you can help with non-parametric linear regression.

PaigeMiller — Sat, 11 Sep 2021 18:10:35 GMT

The data does NOT have to be normally distributed for a linear regression to work. https://blogs.sas.com/content/iml/2018/08/27/on-the-assumptions-and-misconceptions-of-linear-regression.html

The errors around the regression line have to be normally distributed (and independent and common variance).

So go ahead and fit a linear regression and then check the residuals to see if they are approximately normal.

Re: I am wondering if you can help with non-parametric linear regression.

DJGS — Sat, 11 Sep 2021 18:21:00 GMT

WONDERFUL! Thank you.

Re: I am wondering if you can help with non-parametric linear regression.

DJGS — Sat, 11 Sep 2021 18:21:47 GMT

Thank you!!!!