abnormal data, log transform data, still abnormal, what to do?

#1
Hello all,

I have a dataset where I want to run repeated measures and I have abnormal data. So I log transformed the data set and it is still abnormal. I obviously can't run the test with abnormal data so I would like to know if there is anything else I can do to normalize my data. I've looked for outliers and there are some but I was told by my adviser not to remove them because they are physiological outliers. She said I should just do Wilcoxon for time and mann whitney for group. Does anyone else have any other alternative solution?

I am working on SPSS v. 17

thank you in advance

Cherries
 

hlsmith

Omega Contributor
#2
These suggestions are good substitutes for statistics for normally distributed data, but don't address repeated measures. Perhaps you should look into Friedman's Test.
 

Dason

Ambassador to the humans
#3
How did you check for normality? Were you looking at the raw data or the residuals from the analysis?
 
#4
I think friedman's test is exactly what I need.

Raw data. This is how I was taught how to do it.
In spss: analyze: descriptive statistics: explore

Thank you.
 
#5
Sorry guys, I just realized Friedman's test does not give me the option of separating by group (I have 2 groups with 3 time points and I want to see if there is a time and/or group effect). Is there any way around this?
 

hlsmith

Omega Contributor
#6
Might need to just use a generalized linear model (GLM). Check out this topic and let us know if you have any questions.

or

back to Dason's inquiry, normality should be examined via residuals from your model (repeated measures ANOVA). May want to futher investigate this.
 
#7
Hello sorry I just noticed the response. My adviser told me I should use the raw data to check for normality with this data set because my N = 10 & 8 in each group.

However, I do have another data set where my total N = almost 200. I will just be running simple correlations followed by multiple regression. For this, I was advised to use residuals to test for normality, but which ones? I asked a colleague and he suggested I used unstandardized & studentized residuals, but he couldn't explain why...

Does anyone have an explanation for this?

Thank you in advance,

Cherise