topic Testing for independence in ANOVA in Discussions

Testing for independence in ANOVA

nikolaj — Wed, 15 Jan 2014 12:58:37 GMT

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

Re: Testing for independence in ANOVA

louv — Wed, 15 Jan 2014 19:28:56 GMT

ANOVA assumptions
1.) Errors normally distributed>Save residuals and run distribution, normal quantile plot and fit normal>goodness of fit

2.) Variance are the same for all groups> Test for equal variances

3.) Errors occur independently>Plot residuals over time

Re: Testing for independence in ANOVA

nikolaj — Thu, 16 Jan 2014 13:06:27 GMT

Thanks. So for the test of independence in errors i simply make a column from 1 to how many?

Re: Testing for independence in ANOVA

louv — Thu, 16 Jan 2014 14:25:48 GMT

One way to accomplish this is if you are in the fit Y by X platform or fit model you can save your residuals and the under quality and reliability you can create a run chart or even an IR chart to display the residuals over time.

Re: Testing for independence in ANOVA

nikolaj — Tue, 18 Oct 2016 20:32:52 GMT

Here is a screen shot of how it looks when I am in fit y by x. I cannot see quality and reliability anywhere?

Re: Testing for independence in ANOVA

louv — Tue, 18 Oct 2016 20:33:02 GMT

I apologize for the confusion. Once you save your residuals to your data table from the Fit Y by X, you can launch the Analyze>Quality and Process platform and create a run chart and or/IR from there.

Re: Testing for independence in ANOVA

louv — Thu, 16 Jan 2014 16:39:14 GMT

Since I am not exactly clear on your specific problem I would suggest some good references that can easily found by searching google by entering

Testing for Independence of Errors in ANOVA in JMP

Re: Testing for independence in ANOVA

fariasj_hotmail — Wed, 19 Aug 2015 21:43:58 GMT

Hello Lou Valente,

Where test for equality of variances?

Re: Testing for independence in ANOVA

louv — Wed, 19 Oct 2016 00:26:29 GMT

Go to red triangle and select Unequal Variances....

Re: Testing for independence in ANOVA

fariasj_hotmail — Fri, 21 Aug 2015 22:15:20 GMT

Thanks Lou Valente. For mixed effects models have the option to test homogeneity? Or only one way Anova?

Re: Testing for independence in ANOVA

jiancao — Mon, 24 Aug 2015 00:51:47 GMT

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

Re: Testing for independence in ANOVA

David_Burnham — Mon, 24 Aug 2015 12:37:41 GMT

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.