Linear Regression

#1
1) What should a sample size at least be, in order to run a linear regression test? Would 18 be too small?
2) Also, can I perform a linear regression test on data that is left skewed.
3) I ran a linear regression test on data that was normal and all seemed to go well, testing signf. However, the residual plots do not have a flat line about the line y=0). I see a pattern. Does that say my linear equation doesn't predict? Can I still go with my results from my coef. of det value r^2 and other results from my SPSS output when I did the linear regression test? I am not sure how to interpret the non flat pattern result.
Thanks for your help:)
 

hlsmith

Omega Contributor
#2
Please post images of your residuals. If you have only one predictor, and that is appropriate, 18 may be enough data. It comes down to how big of an effect there is and how much variability in data.
 
#3
See the attached file. I hope you don't mind that I sent you an overview of my study and findings. I am concerned that maybe the data count for web (n=18) was too small to do the linear regression test. I am also concerned that I should have not ran the linear regression on the f2f data results b/c it's left skewed. Also, because the p values for my constants (in the coefficients results) for the linear equations (for the web and f2f) are not significant, ...does that mean the other findings from the linear regression test shouldn't be used like correlations and results from the Anova table? Lastly, because the web and f2f do not show approximate flat lines in the residual plot, what does that say about my model? I thank you in advance for your help. Just know that I researched this for hours and hours for many days and right now,..I am stuck:(