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.

Thursday, October 06, 2011

Obtaining variances

As a reminder, to obtain variances by group (for the purpose of checking Fmax, for instance), you will want to follow these steps:

Choose "Data" from the menu across the top of SPSS.
From the Data drop-down menu, choose "Split File."
Click "Organize output by groups," and click over the name of your Grouping (or Independent or Between Subjects or Between Groups) variable.
Click "OK."
This will return you to the main screen, and sort your data into groups, which will each be analyzed separately.

Then choose "Analyze" from the menu across the top of SPSS.
From the Analyze drop-down menu, choose "Descriptive Statistics" and then "Frequencies."
Click over the name of your Outcome (or Response or Dependent) variable.
Uncheck the "Display frequency tables" box.
Click on "Statistics" and then choose "Variance." [While you're here, note that this is also where you would request "skew" or "kurtosis."]
Click "Continue."
Click "OK."

This will send you to an output screen, where your variances for the dependent variable will be listed by each level of the independent variable. From there, it is a simple step to pick out the largest variance and the smallest variance, and calculate Fmax.

Once you have done this, restore your data set to its original, unsplit condition by
Choose "Data" from the menu across the top of SPSS.
From the Data drop-down menu, choose "Split File."
Click "Analyze all cases, do not create groups."
Click "OK."
This will return you to the main screen.

7 Comments:

  • At 4:57 PM, Blogger Shin said…

    When I try to do a oneway ANOVA using only two groups for the homework assignment, one of the tables that results is called "Warnings" and it says "Levene's Test of Equality of Error Variances is not computed because there are less than two nonempty groups."

    What does this mean?

    There are also missing pieces of information on the Test of Between-Subjects Effects table:

    In the "Corrected Model" and "filter_$" rows the Mean Square, F, ans Sig. data are missing.

     
  • At 5:43 PM, Blogger Unknown said…

    So the blog won't let me post a new post even though I'm logged in.

    Questions About Homework

    Normality:
    What graphs/tests do we use to test normality? How and where do we report this in our results sections?

    Equal Variance:
    How and where do we report that equal variance can be assumed (either due to the Lavene's test or the Fmax) this in our results sections? If we used the Fmax, do we need to mention that the Lavene's test was significant?

    Homework Sheet:
    Do we need to explicitly answer all the questions on the homework sheet? (e.g. What similarities and/or differences do you notice in the results of the t tests and of the ANOVA?) If so, where do we answer them in the three seperate results sections?

     
  • At 6:38 PM, Blogger Grace said…

    Hello Shin,

    It seems like the warning message came up because you might have entered the wrong variable in the independent variables box. If you select cases, SPSS automatically makes a new column labeled "filter_$".

    But don't be deceived! The variable you chose for the IV --not the IV (FILTER) -- should still be in the "Fixed Factor(s)" box.

    -Grace

     
  • At 8:33 PM, Blogger Grace said…

    Normality:
    In ANOVA, the DV(s) is/are assumed to be normally distributed. To determine whether the assumption of normality has been violated, you should look at the Q-Q plot and a histogram. Because this is an assumption, you would not report this in the Results section UNLESS it was violated. For the purposes of this homework assignment, you can do it separately from the Results section.

    Equal Variance:
    Once again, because this is an assumption, you would not report it in the Results section UNLESS it was violated. For the purposes of this homework assignment, you can do it separately from the Results section. If you used Fmax, you should report what the value was and why it did or did not violate the assumption of equal variance.

    Homework sheet:
    YES, you absolutely need to answer all the questions on the homework sheet, in addition to the two write-ups (one for the t tests, another for the one-way ANOVA).

    I hope this helps!

     
  • At 10:59 AM, Blogger Mari Clements said…

    You can also get this warning because you forgot to "unsplit" your file...see earlier post on checking homogeneity of variance.

     
  • At 11:03 AM, Blogger Mari Clements said…

    The writeup of the analysis should be similar to the practice example posted on the class moodle site.

    The responses to specific questions should be provided in the order they are asked in the homework and formatted so that each answer is easily seen.

     
  • At 8:47 PM, Anonymous Anonymous said…

    How do you do the Q-Q plot and histogram?

     

Post a Comment

<< Home