Hi!
Im currently trying to do a multivariate meta analysis based on the data from a prospective cohort study with three branches, where each branch has reported three OR:s in total, one for each exposure level. Since the OR:s from each branch are dependent, I have assumed I need to use multivariate MA instead of univariate to account for dependency. I have covariance matrices for each study branch from the original data. I have created a datafile with all variables I believe that I need, but I struggle with the syntax for mvmeta without multilevel metaregression. The ones I find in help files and documents are for mvmeta with multilevel meta regression, which I don't want since I don't have that kind of data. Is it not possible to use mvmeta for meta-analysis of OR:s to just produce a summary table and forest plots? ChatGPT gave me the suggestion to use this code:
"mvmeta lnOR1 lnOR2 lnOR3, se(se_lnOR1 se_lnOR2 se_lnOR3) esvar(mycov_study1 mycov_study2 mycov_study3) study(study_ID) re"
where lnOR1 equals the logOR of lowest exposure level values, lnOR2 - middle, lnOR3 - highest exposure, with their standard errors reported in se_lnOR1 se_lnOR2 se_lnOR3 and esvar being the covariance matrices I manually created in Stata from the values in my data, and study_ID being the identification variable for the different studies.
There's probably plenty wrong in this code. But I cannot even get past the first variable without getting an error message saying "Warning: variable lnOR1 not used (looking for variable names lnOR1+suffix)
No variables found starting with lnOR1". How can this be when I can clearly see the variable in that exact spelling in my datafile? Is it that there needs to be a specific name for that variable for Stata to recognize it when trying mvmeta? Like I said, I have tried to find the syntax but can only find it for multilevel meta-regression, so I'm kind of stuck. So, does anyone know what I have done wrong or how the code is suppose to be written?
Would appreciate anyone who might have some insight that can help a Stata-beginner!
Thanks!
Im currently trying to do a multivariate meta analysis based on the data from a prospective cohort study with three branches, where each branch has reported three OR:s in total, one for each exposure level. Since the OR:s from each branch are dependent, I have assumed I need to use multivariate MA instead of univariate to account for dependency. I have covariance matrices for each study branch from the original data. I have created a datafile with all variables I believe that I need, but I struggle with the syntax for mvmeta without multilevel metaregression. The ones I find in help files and documents are for mvmeta with multilevel meta regression, which I don't want since I don't have that kind of data. Is it not possible to use mvmeta for meta-analysis of OR:s to just produce a summary table and forest plots? ChatGPT gave me the suggestion to use this code:
"mvmeta lnOR1 lnOR2 lnOR3, se(se_lnOR1 se_lnOR2 se_lnOR3) esvar(mycov_study1 mycov_study2 mycov_study3) study(study_ID) re"
where lnOR1 equals the logOR of lowest exposure level values, lnOR2 - middle, lnOR3 - highest exposure, with their standard errors reported in se_lnOR1 se_lnOR2 se_lnOR3 and esvar being the covariance matrices I manually created in Stata from the values in my data, and study_ID being the identification variable for the different studies.
There's probably plenty wrong in this code. But I cannot even get past the first variable without getting an error message saying "Warning: variable lnOR1 not used (looking for variable names lnOR1+suffix)
No variables found starting with lnOR1". How can this be when I can clearly see the variable in that exact spelling in my datafile? Is it that there needs to be a specific name for that variable for Stata to recognize it when trying mvmeta? Like I said, I have tried to find the syntax but can only find it for multilevel meta-regression, so I'm kind of stuck. So, does anyone know what I have done wrong or how the code is suppose to be written?
Would appreciate anyone who might have some insight that can help a Stata-beginner!
Thanks!
Comment