Checking if error in data is normally distributed or not

I have two datasets (x) and (y). How could I know if the error in both data is normally distributed or not? Thank you, Amb

Answer 1

[Star Strider]

One option is the Anderson-Darling test. See the documentation for the adtest function for details. Tare links to others in and at the end of that page.

[amberly hadden]

Thank you star you are always helpful and first in response I appreciate this very much. I'm infect interested to know how could we get an idea if all the data errors are normally distributed ? does this means if data is from a normal distribution data error will also be from a normally distributed?

[Star Strider]

My pleasure.

As I understand your Question, in your example the argument to the adtest function would be the errors (vector of residuals) from a regression or similar procedure. The original data can have any distribution.

[amberly hadden]

Thank you again star. I was reading a paper where the author mentioned "if all the data errors are normally distributed" so now I understand it completely (correct me if I'm still wrong) first, I need to run regression then estimate residual and then have to use adtest on residual to see if errors are normally distributed or not if they are a linear fit is good approximation. I was really confused that how could we know the error in only observations. Thanks again

[Star Strider]

My pleasure.

You are correct. The fit does not have to be linear (you can do a nonlinear fit to your data if that is appropriate). That aside, you would be checking to see if the residuals — the difference between the regression at every value of your independent variable and your dependent variable at those points — are normally distributed.

Answer 2

[Walter Roberson]

kstest() each of them individually.

If you want to know whether they come from the same normal distribution, see kstest2()

[amberly hadden]

thank you Walter... In my understanding, real observations can rarely be from a normal distribution.