Hello,
I am currently working on a meta-analysis project and would like to use the glst and xblc command on multiple studies.
I am running into some programming issues when I try to go through the steps. For some background, in one of my datasets, I have 11 studies, some have 2 exposure levels whereas some have 5 exposure levels, there are case-control studies, cohorts (both IR and CI data) and a dose variable that ranges from approximately 0-10 (with continuous like data to the tenth of a decimal place).
Here are the commands I have followed and their corresponding outputs (in Stata v.13):
n=controls in CC data, person years in CI data and noncases in IR data
case= number of cases
dose= exposure as a continuous variable
logrr= natural log point estimate from each study and exposure level
se= natural log standard error
id= study id in numeric form
study= CC for case-control coded as 1, IR for incidence rate coded as 2 and CI for cumulative incidence coded as 3
mkspline doses = dose , nk (3) cubic
. clist study dose doses1 doses2 in 1/5
study dose doses1 doses2
1. 1 0 0 0
2. 1 .4 .4 .00256
3. 1 1.2 1.2 .06912
4. 1 2.7 2.7 .7724235
5. 1 3.8 3.8 1.856052
. glst logrr doses1 doses2 , se(se) cov( n case ) pfirst( id study )
*: 3200 conformability error
glsest(): - function returned error
<istmt>: - function returned error
r(3200);
end of do-file
r(3200);
. levelsof dose
0 .4 .5 .9 1 1.2 1.5 2 2.1 2.5 2.6 2.7 2.8 3 3.5 3.8 4 4.3 5 6.5 9.9 10 13.5
.
. xblc doses1 doses2, at(`r( levels )') covname( dose ) ref (0) eform line
dose exp(xb) (95% CI)
0 1.00 (1.00-1.00)
value .4 of dose not observed
r(198);
I was wondering if you have ever run into something like this or if you have any ideas to what is going on here.
I really appreciate your time and look forward to hearing from you!
Thank you,
Megan
I am currently working on a meta-analysis project and would like to use the glst and xblc command on multiple studies.
I am running into some programming issues when I try to go through the steps. For some background, in one of my datasets, I have 11 studies, some have 2 exposure levels whereas some have 5 exposure levels, there are case-control studies, cohorts (both IR and CI data) and a dose variable that ranges from approximately 0-10 (with continuous like data to the tenth of a decimal place).
Here are the commands I have followed and their corresponding outputs (in Stata v.13):
n=controls in CC data, person years in CI data and noncases in IR data
case= number of cases
dose= exposure as a continuous variable
logrr= natural log point estimate from each study and exposure level
se= natural log standard error
id= study id in numeric form
study= CC for case-control coded as 1, IR for incidence rate coded as 2 and CI for cumulative incidence coded as 3
mkspline doses = dose , nk (3) cubic
. clist study dose doses1 doses2 in 1/5
study dose doses1 doses2
1. 1 0 0 0
2. 1 .4 .4 .00256
3. 1 1.2 1.2 .06912
4. 1 2.7 2.7 .7724235
5. 1 3.8 3.8 1.856052
. glst logrr doses1 doses2 , se(se) cov( n case ) pfirst( id study )
*: 3200 conformability error
glsest(): - function returned error
<istmt>: - function returned error
r(3200);
end of do-file
r(3200);
. levelsof dose
0 .4 .5 .9 1 1.2 1.5 2 2.1 2.5 2.6 2.7 2.8 3 3.5 3.8 4 4.3 5 6.5 9.9 10 13.5
.
. xblc doses1 doses2, at(`r( levels )') covname( dose ) ref (0) eform line
dose exp(xb) (95% CI)
0 1.00 (1.00-1.00)
value .4 of dose not observed
r(198);
I was wondering if you have ever run into something like this or if you have any ideas to what is going on here.
I really appreciate your time and look forward to hearing from you!
Thank you,
Megan
Comment