One way ANOVAs or Two way ANOVA?

hellokitty

New Member
Hello... I am new here but I am hoping you all can shed some light on a question I have.

I am not sure how to best analyze my data... hopefully I can explain this well.

I am testing behavioural changes in animals as a result of different treatments.

I want to compare the performance before treatment to the performance after treatment for each drug.

For each drug there is also a control group (not trained) and the experimental group (trained)

However, I also want to compare the % change in behaviour as a result of treatment across all drug conditions.

So... should I run one way RM ANOVAs to compare pre- and post-training scores for each drug (this would be one ANOVA for control and one for experimental)...

And then run a one way ANOVA to compare % changes in experimental animals following all drug treatments??

Or should I run a two way rm ANOVA?

Hopefully this makes sense!

THanks so much

bugman

Super Moderator
There may be some confusion here:

One way ANOVA for one factor one response
Two way ANOVA for two factors one response.

Control and experimental are not seperate Factors, they are different "levels" of one factor, say "Group".

Therefore, run it as a ONE WAY Design

P

hellokitty

New Member
Thank you.

My thought is that with the two way the factors would be 1. time (pre test or post test) and 2. condition (drug treatment + control or experimental)

Could I compare everything at once? Or separately?

bugman

Super Moderator
You are right :tup:.

Factor 1 = group (control experimental)
Factor 2 = time (before after)

hellokitty

New Member
Thanks again, but I am still confused, haha.

I was thinking I could do a two way rm and analyze it that way.

But someone told me it is better to run one way rm anovas for each condition (experimental or control)... and then run a one way anova to compare the two (by looking at a change in behaviour)

I am just not sure which is best.

bugman

Super Moderator
But someone told me it is better to run one way rm anovas for each condition (experimental or control)... and then run a one way anova to compare the two (by looking at a change in behaviour)

What is their reasoning for this?

I would run the two way looking for a signifcant interation between group and time. It makes no sense to me to look at these separatley

hellokitty

New Member
That is what I thought originally.

This is what I want to look at:
Drug 1- experimental, drug 1 - control, drug 1- second control
Drug 2- experimental, drug 2 - control, drug 2- second control
Drug 3- experimental, drug 3 - control, drug 3- second control

Each of the above has a pre and post training score

His reasoning was that a 2 way anova will make comparisons that I am not interested in and thus lower my power significantly. Plus, there is no nonparametric alternative if my data does not meet assumptions.

Basically, I want to see if each group shows a change across time (pre vs. post for each) and then compare post-test scores for all experimental groups.

bugman

Super Moderator
It actually looks like a threeway design (Drug; Group and Time).

I think an three factor omnibus ANOVA, followed by planned comparisons of the groups of interest (noted above) will give you much more information than seperate ANOVAS.

Again, it makes no sense to me to go to the trouble of setting up this experiment and then testing your experimental and control groups seperately.