Which Stat analysis to use?

#1
I have a quick question about what kind of analysis to use. I have been out of the stat game for a while and need a refresher. I am doing a research project measuring a parameter on different types of products. For example, I have four products with five data points each. An example data set is below:

Product A - 150, 152, 146, 144, 149
Product B - 143, 142, 138, 145, 136
Product C - 124, 121, 128, 129, 125
Product D - 161, 156, 158, 166, 155

The question is does product A differ significantly from B? from C? from D? Also, is B significantly different from C? from D? etc..

I appreciate any help!
 

gianmarco

TS Contributor
#2
Hi!

I think that it should be clear to you what feature of the data are you going to test: difference in mean or median?

In this case, provided that your samples are independent, I think that Kruskal-Wallis test (non-parametric test working on medians) should work.

On K-W test, see e.g. http://www.businessbookmall.com/IS NPHTOD21.pdf


See attached files; results from SigmaPlot11.

Best Regards,
Gm


p.s.
note that while performing K-W test, SigmaPlot informed me that assumptions for parametric One Way-ANOVA was met.
 

Dason

Ambassador to the humans
#3
If you're looking for differences in means you could make a few extra assumptions and go the ANOVA route as well. It seems like your data doesn't violate any of the assumptions.
 
#4
I agree with the Kruskal-Wallis
ANOVA is what you want, but you only have 5 data points for each so you have to use the non-parametric equivalent

Good luck!
 

Dason

Ambassador to the humans
#5
I agree with the Kruskal-Wallis
ANOVA is what you want, but you only have 5 data points for each so you have to use the non-parametric equivalent

Good luck!
I agree it's a better idea to use the nonparametric with a small sample size. I just like to point out that if you're willing to make the assumptions then there is nothing wrong with the ANOVA and if the assumptions are actually true then it's more powerful.
 
#6
Thanks for the help. The data sets I used was just a random example. I should be supplied the data soon from my technician. I will forward that on to the group as soon as I get it. My first instinct was to use ANOVA but I like the suggestion of the Kruskal-Wallis. We can take a look when I have actual data.