PG851: GLM--ANOVA

This blog is for the use of students enrolled in either section of PG851. It is intended to be a forum for asking, discussing, clarifying, and helping.

Tuesday, January 19, 2010

the questions begin...oneway anova homework

  • May we please have an example of how to write up a results section with more than two groups?
  • On (D.) we are supposed to "add a sentence explaining your findings" to our results section? What does that mean exactly?
  • How do we know if we have met the assumption of inde. of observations?? Do we need a plot or are they just independent because it's two seperate groups w/ each case in only one group?
  • For (E.) regarding normality and outliers, we use histogram, Q-Q plot, and Cook's d...is that correct? And if so, we use them from the ANOVA we just ran, not a regression model? The plots/graphs are a little different depending on which one (model) they come from.
  • And double checking on (F.), we just look at Levene's test correct?

Thank you so much!

Nikki

posted by Nikki Frederick @ 4:55 PM  

5 Comments:

  • At 6:47 PM, Blogger KGL said…

    --I will email Mari about the possibility of another example

    --It seems to me that when asked to explain your findings one should not simply give the statistical results, but describe in particular what this mean, including things such as what was entered into the model as well as direction of effect. (things that you had done in regression, just never explicitly asked for)

    --Independence of observations is just what you described, that observations are part of one and only one group in the model, without outside influence on the other groups. This is more of a theoretical assumption to think about, but no, i do not believe there are tests in SPSS for this.

    --Yes, histogram, qq plots, and cook's d would be ways to test these assumptions. And yes, You are testing these assumptions for your ANOVA not a regression analysis.

    --Yes, Levene's will test homogeneity of variance.

    Hope this helps!

     

  • At 10:34 PM, Blogger Mari said…

    I have every confidence in your ability to extrapolate from the three examples that you were given to a three-group example. Remember also that you have been reading other sources in your article reviews and in your readings from the text that should inform your own writing.

    One key piece to remember is that you cannot use comparative terms in the absence of statistical tests that justify them. That is, you cannot say that one group is different from another without a test that actually compares those two groups. We've gone over several examples of these kinds of tests in class.

     

  • At 10:36 PM, Blogger Mari said…

    When/if you have a significant Levene's test, are you done?

     

  • At 9:29 AM, Blogger Nikki Frederick said…

    Thank you! This all makes sense with what I had assumed, but I wanted to make sure.

    As far as the write up when we have 3 groups, I was just wondering how to word things such as (group 1 was different from group 2, but not from group 3, etc.) I understand the numbers that should be reported, I'm just a little confused on how to report which groups were different from one another.

    Also, reporting CI is optional correct??

     

  • At 8:06 PM, Blogger Mari said…

    As long as you provide the stats to support your conclusion, wording it much like you did in this post is just fine.

    For this assignment, reporting CIs is completely optional. Make sure you give me some indication of an effect size, though.

     

Post a Comment

<< Home