Factorial ANOVA Hw
I keep getting no significant interaction for either Lifequal or Lifequal2. I don't think this is correct, since we are asked to run post hoc tests, but I am not sure what I am doing wrong. I did the practice homework correctly.
I also noticed that I get a warning when I open up the data set that says something like the data set is written in a character encoding (Windows- 1252). I'm not sure if this matters or not.
I also noticed that I get a warning when I open up the data set that says something like the data set is written in a character encoding (Windows- 1252). I'm not sure if this matters or not.
posted by ashleynevin @ 11:09 AM
7 Comments:
At 2:42 PM,
Anonymous said…
I also get no significant interaction for either Lifeequal or Lifeequal 2 and the IVs.
Also, I did not get a statistically significant interaction for these variables.
At 3:21 PM,
Anonymous said…
To Mari and TAs, can the students receive more time to complete the factorial hw that is due this Friday at 5pm? We covered so much new material recently and some of us have questions about the analysis that we run.
At 5:33 PM,
Mari Clements said…
No, you may not have more time. Getting behind at this point in the quarter would only make the end of the quarter horribly painfully busy for you.
Think carefully about under what conditions you need to run post-hocs, and (as the homework instructs) run only those that are necessary.
At 5:37 PM,
Mari Clements said…
BTW, just to correct your phrasing so that you can do it correctly in your writeup--you get an effect OF or FOR the predictors (i.e., a significant interaction effect of Gender and/or Region and/or a significant main effect of Gender and/or a significant main effect of Region) ON the DV (i.e., Quality of Life). You do not get a significant effect FOR the DV.
At 5:55 PM,
Anonymous said…
just to clarify, when you say run only those that are necessary, do you mean, yes, run a post-hoc when a post-hoc is needed, but for whatever that post-hoc may analyze, run BOTH Bonferroni and LSD post-hocs?
At 9:10 PM,
Gemma said…
Is it inappropriate to run the post-hoc from the very beginning with two levels of IV - if one has more than 3 levels, even if the other only has 2? This seems to save time.
At 5:58 AM,
Mari Clements said…
Anonymous--
Yes, you should follow the homework instructions, and when/if a post-hoc is necessary, run both types specified.
Gemma--
A post-hoc is *only* justified when you have a significant main effect. Note that, just as you can have a significant effect without a significant pairwise comparison, you can also get a significant pairwise comparison in the absence of a significant effect. It would take tremendous discipline to ignore that significant post-hoc, but you never should have run it if the omnibus analysis wasn't significant. Thus, conceptually it is not a good idea to run post-hocs until you know you have a significant effect. You should *never* run a post-hoc on any predictor variable that has only two levels.
Post a Comment
<< Home