Dear All,
I am new to MlwiN so I apologize in advance for the naïve question.
Anyone knows where are stored and whether it is possible to extract the individual slopes and intercepts for each subject as estimated by the model examined? If yes, how can I do that?
Many thanks for all you help
Sarah
individual slopes and intercepts
-
ChrisCharlton
- Posts: 1384
- Joined: Mon Oct 19, 2009 10:34 am
Re: individual slopes and intercepts
You should be able to calculate these using the residuals window (see chapter 3 of the user guide: http://www.bris.ac.uk/cmm/software/mlwi ... al-web.pdf). If you are fitting a growth-curve model you might also want to look at section 13.3. If you want to copy these into another package then, once these have been saved into a column, you can highlight them in the Names window and click the Copy button to put them on the clipboard.
Re: individual slopes and intercepts
Many thanks for your kind reply and exhaustive reply/
Do you think this could work even for a model with random intercept and random slope?
If my understanding is correct, I should just compute each person residual and then add it to the fixed part of the parameter estimates to get his own slope/intercept?
Do you think this could work even for a model with random intercept and random slope?
If my understanding is correct, I should just compute each person residual and then add it to the fixed part of the parameter estimates to get his own slope/intercept?
Re: individual slopes and intercepts
I am taking over from Chris in replying.
Although MLwin can be sued in a point and click mode; it assumes you know what you are doing! It is therefore well worth spending a couple of hours working through the examples in the Manual so you get a good feel for the way the software works; before you undertake your own modelling
I am guessing but I presume that you have a two level model with occasions nested within subjects and that you have a response and an occasion variable measured variable like Age. You would typically have a constant ( a set of 1's) to get an overall intercept and a slope for Age in the fixed part of the model - this gives the general change for all subjects. You would then typically add a random intercept so that each subject has their own differential intercept and a differential slope . This would give three parameters at the subject level - variance of the intercepts, variance of the slopes and the covariance - these are the key terms for knowing how between subject variance changes with Age.
If you then wanted to have a look at specific subjects you could as Chris suggested estimate the level 2 residuals which would give you the differential intercepts ( for when the Age variable is zero) and the differential slope.
The Prediction facility in MLwin allows you to to predict the linear modelled trajectory for each subject as a function of the general intercept and slope and the differential intercept and slope.
A detailed discussion of this sort of modelling is the first part of Applied Longitudinal Data Analysis: Modeling Change and Event Occurrence
by Judith D. Singer & John B. Willett , New York: Oxford University Press, March, 2003 which has its own website at http://gseacademic.harvard.edu/~alda/
Very usefully you can get the data used in the book and how to model these data with MLwin ( and lots of other programs) from the UCLA website
http://www.ats.ucla.edu/stat/examples/alda/
Hope this helps -
Although MLwin can be sued in a point and click mode; it assumes you know what you are doing! It is therefore well worth spending a couple of hours working through the examples in the Manual so you get a good feel for the way the software works; before you undertake your own modelling
I am guessing but I presume that you have a two level model with occasions nested within subjects and that you have a response and an occasion variable measured variable like Age. You would typically have a constant ( a set of 1's) to get an overall intercept and a slope for Age in the fixed part of the model - this gives the general change for all subjects. You would then typically add a random intercept so that each subject has their own differential intercept and a differential slope . This would give three parameters at the subject level - variance of the intercepts, variance of the slopes and the covariance - these are the key terms for knowing how between subject variance changes with Age.
If you then wanted to have a look at specific subjects you could as Chris suggested estimate the level 2 residuals which would give you the differential intercepts ( for when the Age variable is zero) and the differential slope.
The Prediction facility in MLwin allows you to to predict the linear modelled trajectory for each subject as a function of the general intercept and slope and the differential intercept and slope.
A detailed discussion of this sort of modelling is the first part of Applied Longitudinal Data Analysis: Modeling Change and Event Occurrence
by Judith D. Singer & John B. Willett , New York: Oxford University Press, March, 2003 which has its own website at http://gseacademic.harvard.edu/~alda/
Very usefully you can get the data used in the book and how to model these data with MLwin ( and lots of other programs) from the UCLA website
http://www.ats.ucla.edu/stat/examples/alda/
Hope this helps -