ANOVA or T-test

true_friend

New Member
Hi
I have two categorical variables as in below example:
Code:
Region	Dim	Value
PK	Dim1	2
PK	Dim2	3
PK	Dim3	-1
PK	Dim4	-4
US	Dim1	-0.5
US	Dim2	4.3
US	Dim3	2.1
US	Dim4	2
PK	Dim1	5
PK	Dim2	-0.4
PK	Dim3	-0.3
PK	Dim4	0.8
US	Dim1	1
US	Dim2	2
US	Dim3	0.8
US	Dim4	1
I want to compare the variables US and PK, if they are different from each other with respect to any or all of the four attributes (Dim1, Dim2, Dim3, Dim4). My question is:
1. Should I use a one way ANOVA and then post hoc (for example HSD Tukey) to determine if these two variables have any significant differences at any of these 4 attributes (PKDim1-USDim1, PKDim2-USDim2...)?
Since I have to report the results in 1 graph (box plot), i thought using an ANOVA would be fine. But, the issue of non-desirable group comparisons (e.g. PKDim1-PKDim2....) is confusing. Is it acceptable to perform ANOVA in such situations?
OR
2. Should i just preform T-test 4 times, for example US versus PK on Dim1, Dim2, Dim3 and Dim4. And report each one separately.