Adjusting for clustering in spss

[Natalie Sappleton]

Hi – I hope someone can help.

I’m analysing the results of a factorial study. Each respondent (n=25) completed 8 vignettes. Total number of observations= 200.  The unit of analysis is the vignette, so I understand I have to adjust for clustering at the participant level to reduce standard errors.

How do I go about this in SPSS?

Thanks in advance

[J. Andres Dominguez Gomez]

Hi. I think I don’t understand very well. If you ask spss a factor analysis with n variables, I think the only think you have to do is to ask it for the Sphericity Test of Barlett, by click the option into the window. If that test is positive (high value), you can relie on your results, that is, you can analyse your results and get confidence conclusions from them. If you have qualitative variables, perhaps is better to do a cluster analysis. Or perhaps, transforming variables in the sense that spss help say to you, comparing results between those two techniques.

Greetings.

[Anonymous]

i might be wrong but i dont thinke Natalie’s referring to a factor analysis here. she’s specifically saying that this is a factorial study which means its a factorial design suitable to some sort of General Linear Model like ANOVA.

 

she very well recognised that this is  a nested design because the respondents are nested within the vignettes, which means her standard (or mean squared, in this case) errors will be biased downwards, inflating the type-1 error rate. the best way to handle this is to use SPSS’s ‘Mixed Model’ under the drop-down menu ‘Analyze’ so that she can get the correct variance components.

[Natalie Sappleton]

Oscar – that is exactly what I was looking for, I think. Thank you so much for your advice.

Natalie

[Author]

Posts