11
$\begingroup$

1) What is the difference between conducting a Linear Mixed Models and an ANOVA?

2) In which circumstances do we conduct a Linear Mixed Models Analysis?

$\endgroup$

1 Answer 1

12
$\begingroup$

1) What is the difference between conducting a Linear Mixed Models and an ANOVA?

ANOVA models have the feature of at least one continuous outcome variable and one of more categorical covariates. Linear mixed models are a family of models that also have a continous outcome variable, one or more random effects and one or more fixed effects (hence the name mixed effects model or just mixed model).

There are sub-classes of ANOVA models that allow for repeated measures, a mixed ANOVA which has one within-subjects (categorical) covariate and at least one between-subjects (categorical) covariate, and repeated measures ANOVA which has at least two within-subjects (categorical) covariate and at least one between-subjects (categorical) covariate.

2) In which circumstances do we conduct a Linear Mixed Models Analysis?

  • when we have a continuous outcome variable
  • when data are clustered (for example, repeated observation on participants or students within classes)
  • when we have sufficient number of clusters to enable estimation of the random effect (variance)
  • when we are not interested in the "effects" of the clusters themselves.

Additionally, ANOVA cannot be used (though there may be a work-around), and mixed models offer a much better alternative, when

  • we have missing data, or
  • the experimental design is unbalanced, or
  • we have multiple (cross-classified or nested) random effects, or
  • we would like to allow the effect of covariates to differ among each level of a grouping variable (random coefficients or random slopes), or
  • when we have an outcome variable that can't be plausibly considered as continous (such as count data and nominal data) - in which case we would use a generalised linear mixed model.

3) How do we obtain such a graph using the above model (Mixed Model or ANOVA) in SPSS to compare the "Low" and "High" condition of the product?

The figure appears to be a simple plot of means for 4 groups. Since it appears to be purely descriptive it isn't therefore something to be obtained from a model.

It appears to be typical of the type of data analysed with a two-way ANOVA - that is, a model with a continuous outcome variable, and two categorical covariates.

$\endgroup$
5
  • $\begingroup$ That helps clarify my question. In my case i am examining the influence of one independent var with 2 levels, onto two dependent var. Do you advise me to use linear mixed methods or MANOVA. It is a within subject test design whereby the same subjects were exposed to the two stimuli (2 llevels of the independent var). kindly advise $\endgroup$
    – user39531
    Sep 10, 2016 at 16:07
  • $\begingroup$ @user39531 it is difficult to say without more information. Please ask a new question with the mixed-model and anova tags, giving as much detail as possible about your experimental design. $\endgroup$ Sep 10, 2016 at 22:08
  • $\begingroup$ I posted a new question under "Which type of Analysis." Thanks $\endgroup$
    – user39531
    Sep 11, 2016 at 1:21
  • $\begingroup$ Just curious how do we do a bar graph to compare the Means of the dependent var in SPSS. I tried graph>legacy>bar> clustered etc, but could not figure out. I have for example the following dependent var: PQ_High and PQ_Low; and ATT_High and ATT_low (as scale interval data). I want to compare the low and high mean of PQ and of ATT as a bar graph. Please advise $\endgroup$
    – user39531
    Sep 11, 2016 at 3:26
  • $\begingroup$ @user39531 questions about how to use software are off-topic here. You could try asking on stackoverflow. $\endgroup$ Sep 11, 2016 at 8:22

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.