Bookmark and Share

Notice: On April 23, 2014, Statalist moved from an email list to a forum, based at

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

Re: st: RE: combining reliability trial data

From   Hilary Taylor <>
Subject   Re: st: RE: combining reliability trial data
Date   Tue, 14 Feb 2012 11:30:25 +0000

Dear Nick,

I used -version 7- with -bagroup- which worked, thanks.

I have decided not to combine my reliability data because it makes my results less comparable with the literature. And to do inter-rater for t2 would be less representative of the way the measure will be used.

I intend to present intra-rater results for both raters (primary and validation trials), and inter-rater results for t1 (primary trial) for both raters, ie 3 Bland & Altman plots.

Thanks for your replies.

Best regards


--On 13 February 2012 18:57 +0000 Nick Cox <> wrote:

The larger question is "to combine the data somehow to make use of it
all" and the answer would be a model that matches your idea of the
data generation process.


On Mon, Feb 13, 2012 at 5:13 PM, Nick Cox <> wrote:
The error in using -bagroup- arises, I guess, because it lacks a
-version- statement. Inserting a line such as

version 7

after each -program- statement is likely to fix the problem that braces
are used within the program in ways that clash with present rules. I
have not tested this guess and there may be other problems.

See also [P] version for details here.

-bagroup- comes from the following place. Please remember that you are
asked to explain _where_ user-written programs you refer to come from.

STB-55  sbe33 . . . . Comparing several methods of measuring the same
quantity        (help baplot, bagroup, bamat, sdpair if installed)
 . . . . .  P. Seed        5/00    pp.2--9; STB Reprints Vol 10,
       commands implementing the Bland-Altman approach to comparing
       two or more measurement methods;  also alternative command to
       sdtest based on Pitman's method giving confidence intervals
       for variance ratios of paired data


Hilary Taylor

I am hoping to test the inter- and intra-rater variance of a measurement
method.  I had two raters (R1 and R2) perform measurements on 28 items
and each rater measured the set twice (t1 and t2).  The measurements
are scale.

I have drawn up some Bland & Altman plots and have performed Pitman's
tests.  I have carried this out for inter-rater (R1t1 vs R2t1) and
intra-rater (R1t1 vs R1t2), however I would really like to combine the
data somehow to make use of it all.  I thought the command 'bagroup'
(Paul Seed, 2000) might be useful but have been unable to explore this
because the following error comes up:

bagroup  t1_r1  t2_r1  t1_r2  t2_r2
program error:  matching close brace not found

I would be grateful if you could tell me firstly, if this would be a
useful thing to do (if I can get it working) and secondly, if you have
any suggestions for a better way for me to combine these data (I would
like to create plots and associated stats)?

(I am using Stata/SE 11.2)

*   For searches and help try:

Hilary Taylor
Research Assistant
Academic Rheumatology
Musculoskeletal Research Unit
School of Clinical Sciences
University of Bristol
Avon Orthopaedic Centre
Southmead Hospital
Bristol. BS10 5NB

Tel: 0117 323 6187

*   For searches and help try:

© Copyright 1996–2018 StataCorp LLC   |   Terms of use   |   Privacy   |   Contact us   |   Site index