we are college students who are currently using SPSS to encode our experiment's data. but somehow, the normality test came out this way and we're not really sure on what else to use except ANOVA and the Shapiro-WIlk test. is there any other tests included within SPSS that we can utilize?
https://www.youtube.com/watch?v=YAlJCEDH2uY&pp=ygUfY2VudHJhbCBsaW1pdCB0aGVvcmVtIHN0YXRxdWVzdA%3D%3D
I had never seen any of the StatQuest videos before, but that one Cristian Ricci pointed to is generally pretty good, I think. I do wish, though, Josh Starmer had said that the sample means are approximately normally distributed rather than normally distributed. Maybe that is included in the "fine print" he alluded to. In any case, I will check out some more of his videos. Thanks for alerting us to his channel, Cristian.
Here is a short conference presentation on normality tests that I gave several years ago that might also be helpful to Juliana Gajudo and her colleagues.
Presentation Silly or Pointless Things People Do When Analyzing Data: 1. ...
Cheers,
Bruce
There are non-parametric Kruskal-Wallis one-way analysis of variance,
it comes out as H.
ANOVA remains a reliable method for comparing means across groups, but alternatives like regression analysis, the Kruskal-Wallis test (for non-parametric data), or MANOVA (when analyzing multiple dependent variables) may be more appropriate depending on data characteristics, research goals, and assumptions such as normality or homogeneity of variances.
If ANOVA assumptions (normality, homoscedasticity) are violated, consider non-parametric alternatives like Kruskal-Wallis, permutation tests, or robust regression techniques.
Guys, non-parametrics are limited by many ways. Moreover, the central limit theorem can be assumed as valid with those sample sizes (~30 by group). Finally, ANOVA and t-test are also very-very robust with respect to assumption violations! I would definitely use a parametric approach (Better a GLM based one, so that you can adjust for a confounder or two). I am aware that in certain fields non-parametric are very popular. Make a supplementary analysis based on non-parametric and keep it on the side.. Again, non-parametric are ok for few situations such as small samples from animal studies or experiments were the samples are very small. A note of statistical power in those situations can be find here Article Determining sample size adequacy for animal model studies in...