No post-hoc tests for Mixed between-within subjects ANOVA aka split-plot ANOVA ?

#1
Hello,

I am performing some "mixed between-within subjects ANOVAs" aka "split-plot ANOVAs" aka "SPANOVAs".

I am following Pallant 2011.

There is no mention of post-hoc tests for the "mixed between-within subjects analysis of variance". There's a link to the exact chapter I'm following here:
http://ebooks.narotama.ac.id/files/...ween-Within Subjects Analysis Of Variance.pdf

Can anyone explain (simply) why this is?

I was just going to follow how to perform it, and then write it up inline with the example write up provided. Is there anything else I should do? I was under the impression from this chapter that if I followed this chapter, this is all I would need to do.

Thank you for your time. :)
 

Miner

TS Contributor
#3
There are a number of reasons why you may not perform a post-hoc test. While not a complete list, these may include:

  • There are only 2 levels, so it would add no additional information
  • The variable of interest is a random variable. Post-hoc is only done for fixed variables
 
#4
Hello,

Thank you for your response.

In this case, with a "mixed between-within subjects ANOVA" aka "split-plot ANOVA" aka "SPANOVA", with one IV as a group, and another as time, both with only two levels (two groups, and two time points), would this mean there are only 2 levels?

Could you provide a text source/reference/citation for "There are only 2 levels, so it would add no additional information"?

Thank you for your time/assistance. :)
 

Miner

TS Contributor
#5
Yes, you only have two levels for each IV. ANOVA is an "effects" test, while post-hoc tests are typically "means" tests. From a practical perspective the distinction is that an effects test can only state that at least one of a group of means is different, but cannot tell which mean it is. That is why you perform a post-hoc test, to identify which specific means are different. However, when the ANOVA is performed on an IV with two levels, there is no question about which means are different, there are only two means. Hence, no additional information is provided by a post-hoc test.


See the second sentence in this handout: http://pages.uoregon.edu/stevensj/posthoc.pdf