Testing for independence in ANOVA

[#1]

Hi

I am conducting an ANOVA and I am in the process of performing the assumption checks.

I have checked for normality and homogeneity of variance. However, I am not sure about how

to create a scatter plot in order to test for independence of errors.

In my research design respondents have been divided on the basis of what type of coke they

drink (diet or regular), on a colour basis (red or green), and on an information basis (little or much).

They have for instance answered a question about the attractiveness of the product under study

on the basis of a 5-point likert scale. With this information then how do I test for independence of

errors in SAS JMP?

Thanks!

Nikolaj

[#2]

Unlike OLS linear regression mixed models assume neither independence nor homogeneous in errors. After you fit a mixed model with a certain covariance structure (e.g, UN or AR (1), etc) you can determine whether a variance or covariance estimate is statistically significant by checking its 95 confidence limits or conducting a z-test. See my blog post for details. JMP Pro for linear mixed models — Part 1

[#3]

As Lou said, to test for independence of errors you can plot the residuals.  If you want to be more formal in testing for independence you can test for first order autocorrelation by using the Durbin-Watson test available in the Fit Model platform (red triangle>Row Diagnostics).  Note that by default this test gives you a test statistic - you need to go to the associated hotspot to see the associated significance level.  A significant result implies that the data are not independent.