Review Session Questions
1) Is Levene's test only looked at if cell size ratios are greater than 4:1, or is it always used?
Levene's is always looked at as a preliminary step. If it's significant, always follow up with an Fmax test.
Levene's is always looked at as a preliminary step. If it's significant, always follow up with an Fmax test.
2) Which row for Levene's test should be used: equal variances assumed or not assumed?
I think there is some confusion here. There's only one line in the Levene's test. The equal variances assumed and equal variances not assumed lines are for the t values in a t test, and the Levene's test tells you which of those lines to examine.
I think there is some confusion here. There's only one line in the Levene's test. The equal variances assumed and equal variances not assumed lines are for the t values in a t test, and the Levene's test tells you which of those lines to examine.
3) What is the APA standard for reporting a p value of .0498? p < .05? p = .05?
The only time < is used is when p < .001. This would be an unfortunate p value to obtain, and if one did obtain it, one might consider whether or not putting the values in a table were possible.
The only time < is used is when p < .001. This would be an unfortunate p value to obtain, and if one did obtain it, one might consider whether or not putting the values in a table were possible.
4) Should they know a quantitative way of determining outliers? (p. 89 of the text includes a general rule of thumb based on z)
Not necessary. We will get to box & whisker plots, which are probably the best graphic way to identify outliers, but for now, just think about what an outlier might look like on a histogram or a Q-Q plot. (But great that you are reading the text that carefully!)
Not necessary. We will get to box & whisker plots, which are probably the best graphic way to identify outliers, but for now, just think about what an outlier might look like on a histogram or a Q-Q plot. (But great that you are reading the text that carefully!)
5) Could they have a few examples of ANOVA write-ups on moodle?
Would you settle for one, for now?
Would you settle for one, for now?
posted by Mari Clements @ 5:06 PM
3 Comments:
At 8:42 PM,
Anonymous said…
I need a little clarification on the Levene's test. I have in my class notes that a Levene's of < .05 is good BUT in the notes from the review session I have that we want a Levene's of > .05 and that a significant Levene's test is a bad sign for research. Basically I'm confused. I thought that a low number on a Levene's test was good and meant you could read the"Equal variance NOT assumed" line (meaning you can reject the Null Hypothesis).
At 11:25 AM,
David said…
We do not want Levene's test to be significant when testing for homogeneity of variance. If you recall, the null hypothesis is that there is no difference between groups. A significant Levene's test would indicate that there in fact is a difference in variance between the groups (which we don't want). It's a bit confusing because in our research, we want to see significant results.
Hope that helps.
At 12:46 PM,
Mari Clements said…
Anonymous wrote:
I have in my class notes that a Levene's of < .05 is good.... I thought that a low number on a Levene's test was good and meant you could read the"Equal variance NOT assumed" line (meaning you can reject the Null Hypothesis)
---
A Levene's with a p of < .05 means that you must reject the null hypothesis that you have homogeneity of variance (i.e., no difference between the variances of the groups).
ANOVA and t tests assume homogeneity of variance, and a significant Levene's test tells you that that assumption is NOT warranted.
With ANOVA--and ANOVA ONLY--you would follow up a significant Levene's test with an Fmax test. The Fmax test will tell you whether or not you can proceed with the ANOVA without adjusting alpha. (Because ANOVA is robust to the assumption of homogeneity of variance, and so will still yield usable results even when the assumption is violated, as long as the violation is not too severe. Fmax will tell you this.)
Post a Comment
<< Home