Re: st: Bland Altman Plots for Clustered Data

[David Airey <david.airey@Vanderbilt.Edu>]

There is a Berkley Press statistics paper on BA problems with  
replicated data you should google for... Don't have the reference  
handy, sorry.

Sent from my iPhone

On Apr 17, 2009, at 1:56 PM, "Ploutz-Snyder, Robert (JSC-SK)[USRA]" <robert.ploutz-snyder-1@nasa.gov 
> wrote:

> Statalisters;
> In response to a reviewer's suggestion, I'm creating some Bland  
> Altman plots (using concord.ado) on a panel dataset, where two  
> parallel observations are made on subjects repeatedly over different  
> speeds at a task. I'm doing this for a few outcomes, but have shown  
> a sample below from the first 2 subjects:
>     +---------------------------------------+
>     | subject   speed       rer0       rer1 |
>     |---------------------------------------|
>  1. |       1       0   .9453608   .8185958 |
>  2. |       1       2   .7703036   .7761767 |
>  3. |       1       3   .7817145   .8262445 |
>  4. |       1       4   .8781749   .8886828 |
>  5. |       1       5   .9401777   1.012666 |
>  6. |       1       6   .9988574   1.080287 |
>     |---------------------------------------|
>  7. |       2       0   .9489434   .8451807 |
>  8. |       2       2   .8366485   .7932663 |
>  9. |       2       3   .8556526   .8569991 |
> 10. |       2       4   .8677133   .8986406 |
> 11. |       2       5    .946165   .9572051 |
> 12. |       2       6   .9317179   .9195401 |
>     |---------------------------------------|
>
> I don't normally use BA plots, but for now I'm trying to appease a  
> pesky reviewer.  My question is targeted to those of you who use or  
> have an interested in the BA plots...  Since my data are clustered  
> on subject, what's the best approach to plot BA-like associations  
> between means and diffs on the two outcomes? I'm trying to limit the  
> number of BA plots, as the paper already has better plots showing  
> the effects of speed (and other vars) on these two repeated  
> observations. But I have a reviewer who's really stuck on the BA plot.
>
> I'm starting with:
>
> concord rer0 rer1, loa(regline noref)
>
> But troubled by the clustering of observations within subject.
>
>
> Ideas?
> Rob
>
> *
> *   For searches and help try:
> *   http://www.stata.com/help.cgi?search
> *   http://www.stata.com/support/statalist/faq
> *   http://www.ats.ucla.edu/stat/stata/
*
*   For searches and help try:
*   http://www.stata.com/help.cgi?search
*   http://www.stata.com/support/statalist/faq
*   http://www.ats.ucla.edu/stat/stata/