# Which statistic test to use?

#### nsnsns

##### New Member
I will like to make a statistic analysis on some data but do not know which test to use.
My data is listed below (I have 3 treatments and one numerical data column (%), each combination is made in triplicates):
G A I 30,9000
G A I 31,2000
G A I 41,9000
G A II 17,0000
G A II 18,4000
G A II 19,2000
G A III 19,5000
G A III 20,0000
G A III 23,0000
G B I 26,6000
G B I 30,7000
G B I 42,9000
G B II 35,3000
G B II 28,4000
G B II 41,5000
G B III 33,8000
G B III 21,0000
G B III 36,0000
G C I 32,1000
G C I 42,9000
G C I 42,2000
G C II 19,1000
G C II 15,0000
G C II 22,0000
G C III 17,0000
G C III 20,0000
G C III 20,1000
F D I 10,7000
F D I 14,0000
F D I 5,0000
F D II 34,7000
F D II 34,1000
F D II 33,9000
F D III 24,9000
F D III 26,6000
F D III 17,1000
I will like to test whether there is any significant difference between the data values (%) and select the combination (or combinations if there are no significant difference between some of the highest observations) of treatments that result in the highest data value (%).
Hope that someone will help

#### Miner

##### TS Contributor
I would try ANOVA or GLM. If you have reason to believe (based on a theory) that specific combinations would be higher than a control, you could test specific planned contrasts.

#### nsnsns

##### New Member
Tank you for replying. Would you make a 1-, 2- or 3-way ANOVA? What I-am interested in is the best treatment combination, the one that gives the highest data value (it is % valkues), ex GCI is significant, or GCI and GAI is significant higher that the rest.

#### Miner

##### TS Contributor
A 3-way ANOVA or GLM would be appropriate.

I took a quick look at the data. The main effects do not appear to be significant, but there are indications that some of the interactions are significant. You might create a new variable with a level for each unique combination of the IVs and test it using a 1-way ANOVA followed by a post-hoc test such as Tukey's.