Modern repeated measures analysis using mixed models in SPSS (2)

This uses a Repeated measures analyse as an introduction to the Mixed models (random effects) option in SPSS.
Demonstrates different Covariance matrix types & how to use the Likelihood ratio test to evaluate different models.
First an inappropriate standard regression model is developed then one with a random intercept ( considering the patient a level 2 variable) and finally a random intercept+ slope model, each is evaluated using the likelihood ratio test (see previous video for more details on obtaining chi square p value).
The example is from Twisks excellent book - applied multilevel analysis p.91-95 with him using the free package MLWin
Robin Beaumont
for Full notes, MCQ's etc see:
www.robin-beaum...

[@Pietromans]

4:20 to 4:32: isn't it just the opposite? If the independent variable (lifestyle) goes up by one, the dependent variable (health) goes up by 0,14?

[@kangma4462]

Thanks so much for the video. While i noticed the Unstructured correlation between Intercept and Slopes, May i ask what's the evidence for this correlation?

[@jocelynarbuthnott3221]

Hi thank you! I have found this very helpful. Do you know if there is anywhere that demonstrates how to distinguish between and within effects? you mentioned there are ways but you would not go into it for this video, unfortunately have so far not found anything to demonstrate how this is done on spss?...

[@nikkijahanfar]

Hi Robin, I really enjoyed this presentation. I wonder if we can use mixed models for twin data as twins as clusters. Or do you prefer the GEE model. Can you please let me know which one is suitable when we are interested in individual outcomes, but we are concerned with within-twin correlation. Thanks

[@RobinBeaumont]

A difficult question - to really understand what is going on you need to understand the matrix algebra underneath it basically I suggest you look carefully at the pdf handouts that go along with this video and also the next pdf concerned with 5 repeated measures in 2 treatment groups page 18 you can see a link to it from the link mentioned in the video title

[@TheHotHamster]

Hey Robin, sorry I am not an expert, but at 4:32 you say that the parameter of lifestyle .14 means that when you have a one unit change in health you have a .14 change in lifestyle. But isn't it the other way around? Isn't it the unstandardized regression coefficient (slope) which indicates how much the estimated regression line increases at the criterion level when the predictor changes with one unit? Would very much appreciate an answer :) Lisa

[@TheHotHamster]

Ok, I am the second one who comments about that.. didn't see that ;)

[@RobinBeaumont]

Thanks for the comment Pieter will check when I get a moment

[@petsol]

I'd like to know that as well: I thought the between effect is in fact "sucked up" by the random intercept, but the video tells it hasnt. Maybe modelling with a subject average and a RM diff from that subj average would help to model distinctively, but I thought that is done automaticaly by SPSS.

[@angelomark6653]

I get an error that the number of observations exceeds of less than the number of parameters. What exactly does this mean?

[@BreakTheRules2011]

I believe that this means that you have exhausted your degrees of freedom.

[@rebeccadcruz4356]

Fantastic video - thank you! Very helpful

[@RobinBeaumont]

Factor or covariates - Many people have problems with the factor/covariate box here, you are in good company see andy field's Discovering statistics using SPSS 3rd ed. page 744

[@Tomahawk1999]

hello sir, how do you decide which variable goes in factors and which in covariates? could you please help me understand? thanks a lot

[@ArcadianGenesis]

1) Are you saying "SPSSS"? Why the fourth S? 2) That is a really noisy mouse.

[@arucardmaster]

Anyone can speak American English that can teach us?

[@minxialuo498]

Thanks for the explanation! It's good!