Glst command, dose-response meta-analysis

[Rachel Kimble]

Hi, new to Stata.

Can someone please help with glst command the data returns conformability error. It seems to arise from a single study:

glst logrr dose if(id==2), se(se) cov(n cases) ir
*: 3200 conformability error
glsest(): - function returned error
<istmt>: - function returned error

Data for the study is:

id rr lrr urr dose cat n cases type logrr se

2 1 1 1 1.9 0 167657 863 2 0 0
2 1.04 .94 1.14 4.5 1 171701 833 2 .03922071 .04921102
2 1 .9 1.1 7.8 2 173427 793 2 0 .05119244
2 1.04 .94 1.14 13.7 3 173317 814 2 .03922071 .04921102
2 .97 .87 1.07 26.3 4 170814 743 2 -.03045921 .05278687


Can anyone help or point me in the right direction?

Many Thanks,
Rachel

[Andrea Discacciati]

Hello Rachel, the problem is that one non-referent logRR=0 (obs. #3), which bothers -glst- (user-written command from SSC/SJ)

. clear

. input id rr lrr urr dose cat n cases type logrr se

            id         rr        lrr        urr       dose        cat          n      cases       type      logrr         se
  1.
. 2 1 1 1 1.9 0 167657 863 2 0 0
  2. 2 1.04 .94 1.14 4.5 1 171701 833 2 .03922071 .04921102
  3. 2 1 .9 1.1 7.8 2 173427 793 2 0 .05119244
  4. 2 1.04 .94 1.14 13.7 3 173317 814 2 .03922071 .04921102
  5. 2 .97 .87 1.07 26.3 4 170814 743 2 -.03045921 .05278687
  6. end

.
. replace logrr = 0.0001 in 3
(1 real change made)

. glst logrr dose if(id==2), se(se) cov(n cases) ir

Generalized least-squares regression                 Number of obs   =       4
Goodness-of-fit chi2(3)    =    2.27                 Model chi2(1)   =    0.55
Prob > chi2                =  0.5191                 Prob > chi2     =  0.4581
------------------------------------------------------------------------------
       logrr |      Coef.   Std. Err.      z    P>|z|     [95% Conf. Interval]
-------------+----------------------------------------------------------------
        dose |  -.0013552   .0018262    -0.74   0.458    -.0049345    .0022242
------------------------------------------------------------------------------