Solved – Generalized estimating equations output in SPSS

categorical datageneralized-estimating-equationslogisticspss

I am hoping to confirm that I have a suitable way to analyse the different proportions of people who are categorized as left lateralised on the one hand, or bilateral/right lateralised on the other in two different tasks.

I cannot use an ordinary logistic regression (or chi square test) as the conditions are repeated measures.

I have used the Generalized Estimating Equations option in SPSS to allow for the within subjects individual intercepts to vary as for repeated measures, but am wondering how best to interpret the output to show that the proportion of those in each category differs between the two tasks.

Is it correct that I need to re-estimate the model but remove condition from my fixed effects and then compare measures of model fit in each estimation for a significant difference (e.g.using the AIC)? Or should I just stick to looking for a significant Wald statistic and leave it at that?

My final question is how to interpret the pairwise comparisons from this type of analysis. This shows a significant difference between my two conditions, but as I understand it to be a log odds value I'm not quite sure where to go with it.

I do apologise if I haven't made my problems clear, I am quite new to this.

Best Answer

It is generally understood that likelihood ratio tests have better statistical properties than Wald tests. (Edited:) However, as @Macro reminds me, the generalized estimating equations are not a form of maximum likelihood estimation, thus likelihood ratio tests are not available. So you can go ahead with the Wald test that is reported.

It is true that betas are log odds, however, you can exponentiate them and then interpret the result as an odds ratio. If odds ratios aren't sufficiently intuitive (in my experience, people aren't born with the ability to think in odds ratios, but you can learn to use them), you can solve for two cases that have covariate values that seem typical, or are of interest to you, and that are identical except that in the first case CONDITION=0 and the other CONDITION=1. Exponentiating both will yield the two odds; computing $odds/(odds+1)$ in each case will yield two probabilities. Remember that these probabilities and their difference hold only for that exact combination of covariate values. Thus, if you want to know about what happens with a different set of covariate values, you have to go through the process again.

One last point about the interpretation of a model fit by the GEE: this model will describe how the population as a whole behaves, not how an individual within that population will behave. For example, consider a study that looks at students within a classroom taking (and possibly passing) a test. When the model is fit with GEE it is telling you about the class, if it had been fit with a GLiMM instead, it would have told you about an individual student conditional on that student's attributes.

Related Solutions

Solved – Understanding repeated measure ANOVA assumptions for correct interpretation of SPSS output

Your dependent variables should be normal in each cell of between-subject design. You have 2 such cells: 2 groups, so normality should be in both groups. Also, variance-covariance matrix between your 3 DV should be same in the 2 groups. You could check normality by Shapiro-Wilk test or Kolmogorov-Smirnov (with Lilliefors correction) test in EXPLORE procedure. Variance-covariance homogeneity could be tested by Box's M test (found in Discriminant analysis). Note however that ANOVA is quite robust to violations to both assumptions.
Mauchly's test checks the so called sphericity assumption which is necessary for univariate approach to repeated measures ANOVA. This assumption requires that, roughly speaking, differences between your repeated measure DVs don't intercorrelate. If the assumption is violated you should disregard "Spericity assumed" in Tests of Within-Subjects Effects table - there found some corrections (such as Greenhouse-Geisser) instead.
While Tests of Within-Subjects Effects table reflects "univariate approach" in RM-ANOVA, Multivariate Tests table reflects "multivariate approach". These two are both useful and there's a little debate which is "better". Read a little here about them, a bit more here.
Usually one won't check pairwise tests if overall effect is non-significant, it has little sense.

Solved – Interpreting SPSS mixed linear model output

(I don't know if it is proper to answer old questions, especially with a shaky answer, but here goes.)

I'm only learning mixed models myself, but I think you are not doing anything wrong. The parameter estimates are simply calculated in a way that one level of a categorical variable becomes the reference point to which every other level is compared. I think there is a way to change the reference point (at least change from last to first, not sure if you can pick it freely), but in case you want to do multiple comparisons between levels, the answer is not to change the reference point, but to do separate pairwise comparisons. In your case, it would mean adding the following row to the syntax:

/EMMEANS=TABLES(genotype*medication) COMPARE(genotype)

(or perhaps COMPARE(medication), if you prefer them shown in another order).

With EMMEANS you see the estimates in a more natural form (and not in form of effects relative to intercept) and can do pairwise comparisons between individual levels. However, if there are many comparisons, you should remember to do some kind of correction, possibly with a false discovery rate procedure (http://en.wikipedia.org/wiki/False_discovery_rate).

Best Answer

Related Solutions

Solved – Understanding repeated measure ANOVA assumptions for correct interpretation of SPSS output

Solved – Interpreting SPSS mixed linear model output

Related Question