Notice: On April 23, 2014, Statalist moved from an email list to a forum, based at statalist.org.
[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
st: meologit and comparisons
From
Janet Hill <[email protected]>
To
"[email protected]" <[email protected]>
Subject
st: meologit and comparisons
Date
Sun, 1 Sep 2013 17:34:55 +0100 (BST)
Hello,
I have been asked to look at a dataset which consists of 3 intervention procedures, each subject has one of the interventions and they are then measured at 3 time periods, the independent variable is an integer score in the range 1:6. The intervention groups are different sizes and there is some missing data. The object was to see if the interventions and time had any effect. I thought that meologit would be appropriate for this:
. meologit v inter##t || id:, coeflegend
Mixed-effects ologit regression Number of obs = 462
Group variable: id Number of groups = 160
Obs per group: min = 1
avg = 2.9
max = 3
Integration method: mvaghermite Integration points = 7
Wald chi2(8) = 25.93
Log likelihood = -699.97876 Prob > chi2 = 0.0011
------------------------------------------------------------------------------
v | Coef. Legend
-------------+----------------------------------------------------------------
inter |
Control | 0 _b[v:0b.inter]
DVD | 1.862301 _b[v:1.inter]
PPT | 1.95872 _b[v:2.inter]
|
t |
1 | 0 _b[v:1b.t]
2 | 2.75e-16 _b[v:2.t]
3 | -.0267118 _b[v:3.t]
|
inter#t |
DVD#2 | -.4861785 _b[v:1.inter#2.t]
DVD#3 | -.641388 _b[v:1.inter#3.t]
PPT#2 | -1.400277 _b[v:2.inter#2.t]
PPT#3 | -1.04499 _b[v:2.inter#3.t]
-------------+----------------------------------------------------------------
/cut1 | -2.006599 _b[cut1:_cons]
/cut2 | .2863222 _b[cut2:_cons]
/cut3 | 1.172534 _b[cut3:_cons]
/cut4 | 3.385625 _b[cut4:_cons]
/cut5 | 5.400076 _b[cut5:_cons]
-------------+----------------------------------------------------------------
id |
var(_cons)| 5.296882 _b[var(_cons[id]):_cons]
------------------------------------------------------------------------------
LR test vs. ologit regression: chibar2(01) = 133.57 Prob>=chibar2 = 0.0000
I then wanted to compare the interventions and times but I ran into the problems with margins described in http://www.stata.com/statalist/archive/2013-08/msg00445.html. However I found that both pwcompare and lincom appear to work but give different results:
. pwcompare inter, eff
Pairwise comparisons of marginal linear predictions
Margins : asbalanced
---------------------------------------------------------------------------------
| Unadjusted Unadjusted
| Contrast Std. Err. z P>|z| [95% Conf. Interval]
----------------+----------------------------------------------------------------
v |
inter |
DVD vs Control | 1.486445 .5114061 2.91 0.004 .484108 2.488783
PPT vs Control | 1.143631 .5056487 2.26 0.024 .1525774 2.134684
PPT vs DVD | -.3428148 .5056207 -0.68 0.498 -1.333813 .6481835
---------------------------------------------------------------------------------
. pwcompare inter, eff asobserved
Pairwise comparisons of marginal linear predictions
Margins : asobserved
---------------------------------------------------------------------------------
| Unadjusted Unadjusted
| Contrast Std. Err. z P>|z| [95% Conf. Interval]
----------------+----------------------------------------------------------------
v |
inter |
DVD vs Control | 1.493342 .5113274 2.92 0.003 .491159 2.495526
PPT vs Control | 1.149602 .50546 2.27 0.023 .1589188 2.140285
PPT vs DVD | -.3437402 .5053265 -0.68 0.496 -1.334162 .6466815
---------------------------------------------------------------------------------
. lincom _b[v:0b.inter] - _b[v:1.inter]
( 1) [v]0b.inter - [v]1.inter = 0
------------------------------------------------------------------------------
v | Coef. Std. Err. z P>|z| [95% Conf. Interval]
-------------+----------------------------------------------------------------
(1) | -1.862301 .5929079 -3.14 0.002 -3.024379 -.7002228
------------------------------------------------------------------------------
. lincom _b[v:0b.inter] - _b[v:2.inter]
( 1) [v]0b.inter - [v]2.inter = 0
------------------------------------------------------------------------------
v | Coef. Std. Err. z P>|z| [95% Conf. Interval]
-------------+----------------------------------------------------------------
(1) | -1.95872 .5913064 -3.31 0.001 -3.117659 -.7997805
------------------------------------------------------------------------------
. lincom _b[v:1.inter] - _b[v:2.inter]
( 1) [v]1.inter - [v]2.inter = 0
------------------------------------------------------------------------------
v | Coef. Std. Err. z P>|z| [95% Conf. Interval]
-------------+----------------------------------------------------------------
(1) | -.0964189 .5871835 -0.16 0.870 -1.247277 1.05444
------------------------------------------------------------------------------
My questions are
1. Is the use of pwcompare or lincom appropriate
2. Is there an alternative / better way of analysing this data.
Any advice gratefully received.
Stata 13.0
Thank you,
Janet
*
* For searches and help try:
* http://www.stata.com/help.cgi?search
* http://www.stata.com/support/faqs/resources/statalist-faq/
* http://www.ats.ucla.edu/stat/stata/