Using SPSS, I've studied linear regression between two continous variables (having 53 values each), I've got a p-value of 0.000 which means no normal distribution, should I use another type of regression?
Robust regression, generalized models, quantile regression and nonparametric regression
Also you can use transformation that make your response variable closer to normality.
Mohamed Amine Ferradji , you might clarify your question. You got a p-value of 0.000 for what test ? And to what did you apply this test ?
I had the same question as Sal Mangiafico. Perhaps you could also tell us what the two variables are, and clarify what you mean when you say they have 53 values each. Does that mean you have 53 pairs of scores? Or does it mean something else? Thanks.
PS- A scatter-plot might be helpful too!
Consider using the robust least squares or M-estimation and nonparametric regression techniques like local regression (LOESS) or spline regression
If non normal variable is there then you can transform the said variable in to another form in order to get normality otherwise you can go for nonparametric regression models however the accuracy of non parametric regression is less than parametric one so transformation of variable is little bit more better option.
Sal Mangiafico both normality tests (kolmogorov-smirnov and Shapiro–Wilk) has a p-value of 0.000
- Badges
- Science topic
- Similar topics
- Filing