Dear Statalists,
This is a post from a Statalist newbie.
We are performing a diagnostic accuracy meta-analysis with metandi. The problem we are facing is that all of the included studies have no false negative cases, so a warning arises as “Initial values not feasible”. We have also tried adding “force” and/or “gllamm” to the metandi syntax and review the metandi help file, but we weren’t lucky with it.
We are aware that some type of continuity correction could overcome this problem (e.g. just adding 0.5 or 1 false negative case to the last study).
Does anyone know a smarter solution with Stata, avoiding the use of the continuity correction strategy?
Those are our data:
Thanks a lot for your time!
Inigo Gorostiza
This is a post from a Statalist newbie.
We are performing a diagnostic accuracy meta-analysis with metandi. The problem we are facing is that all of the included studies have no false negative cases, so a warning arises as “Initial values not feasible”. We have also tried adding “force” and/or “gllamm” to the metandi syntax and review the metandi help file, but we weren’t lucky with it.
We are aware that some type of continuity correction could overcome this problem (e.g. just adding 0.5 or 1 false negative case to the last study).
Does anyone know a smarter solution with Stata, avoiding the use of the continuity correction strategy?
Those are our data:
| author | tp | fn | tn | fp |
| Minnesota | 4 | 0 | 378257 | 11 |
| Lund | 3 | 0 | 504037 | 9 |
| Schulze | 1 | 0 | 249967 | 31 |
| Frazier | 12 | 0 | 239409 | 1 |
| Zytkovicz | 1 | 0 | 163996 | 3 |
| Feuchtbaum | 1 | 0 | 353891 | 1 |
| Merritt | 107 | 0 | 2802262 | 112 |
Inigo Gorostiza
