Bookmark and Share

Notice: On March 31, it was announced that Statalist is moving from an email list to a forum. The old list will shut down on April 23, and its replacement, is already up and running.

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: st: Correlation between 2 variables overtime- accounting for repeated measures

From   megan rossi <>
To   "" <>
Subject   Re: st: Correlation between 2 variables overtime- accounting for repeated measures
Date   Sun, 17 Mar 2013 00:13:18 +1000

I was thinking of going with xtgee or xtmixed but wasn't quite sure if one would be more appropriate then the other? ( I know gee is more relaxed with equal variance and normality assumptions). 
With respect to the correlation option I didn't mention that the patients had a uninephrectomy following baseline so I think that disturbs the assumption associated with the ar1 option ie. the association with time is not as straight forward . Do you think given this the unstructured option would be best?

Thanks I will look into SEM too....not sure how that would work, if you know of any websites with an example that would be much appreciated!


Sent from my iPhone

On 16/03/2013, at 11:46 PM, "JVerkuilen (Gmail)" <> wrote:

> On Sat, Mar 16, 2013 at 8:32 AM, megan rossi <> wrote:
>> Hi All
>> Can you please recommend what syntax would be most appropriate for my below senario
>> 40 participants with three repeated measures (baseline, year 1, year 2)- observational study ie. no intervention
>> At each of these three time points two continous variables (a) and (b) were measured. I want to know whether (a) and (b) are correlated. If I do a correlation at one time point ie. year 1 the correlation is not significant which I believe is due to the small numbers ie. 40. If I can find a method of accounting for the lack of dependence among these three time points ie. repeated measures I will effectively have 120 pieces of data, which should be sufficient to see a correlation if one really exists.
>> Cheers,
> You could approach this a few different ways. One might be to set up
> the appropriate SEM with correlated residuals to take the longitudinal
> dependence into account. The other would be to set the data up as a
> linear mixed model and use say, AR(1) residuals. I think the SEM
> approach is probably the most straightforward. That said, before you
> do either, be sure to scatterplot both variables broken out over time.
> SEM might wring some extra statistical efficiency, but don't get too
> optimistic.
> *
> *   For searches and help try:
> *
> *
> *

*   For searches and help try:

© Copyright 1996–2015 StataCorp LP   |   Terms of use   |   Privacy   |   Contact us   |   Site index