I have been using STATA 14 for so long and I have run a lot of meta-analysis with no trouble. Recently upgraded and with STATA 17, you can pre-set your Meta data, by Meta set.

The log of my HRs and Lower and upper CI 95% has been the way I do it, Metan is working absolutely fine in the new STATA, Issue is when I tried to do the meta set. I faced the issue of CI asymmetry, which has been discussed here before. I tried the suggested options. But I still find that the forestplots generated using the Std error and HR not the CIs even after easing the CI tolerance.

the computed CIs from the std error is slightly different! I don't like that!

Tried so far:
1)
gen double loghr = ln(hr)
gen double loglci = ln(lci)
gen double loguci = ln(uci)

2)meta set ln_HR ln_Cilower ln_CIHigher, civartolerance(1e-1)

I need to understand why the meta forestplot is utilizing the std error even though it accepts the CIs with the civartolerance(1e-1)?

I like the way the forest plots gets generated when I use the Meta command, if anyone can provide or show me how can I generate the same forest plots settings with metan command would be great

Also and how can I get Meta forest plots to use the actual CIs


This is a table showing the difference in the populated log of CIs and ones generated by metan

+------------------------+
| _meta_cil ln_Cilo~r |
|------------------------|
1. | .35002375 .3506569 |
2. | -.1838624 -.2231435 |
3. | -.66627276 -.6931472 |
4. | 1.4064229 1.406097 |
+------------------------+

and this proves that the STATA accepts my data meta set, but then would still use Std error ! Please see attached forest plots generated by the two ways and the slight difference in CIs
meta set ln_HR _LCI _UCI, civartolerance(1e-1) random(dlaird)

Meta-analysis setting information

Study information
No. of studies: 4
Study label: Generic
Study size: N/A

Effect size
Type: <generic>
Label: Effect size
Variable: ln_HR

Precision
Std. err.: _meta_se
CI: [_meta_cil, _meta_ciu]
CI level: 95%, controlled by level()
User CI: [_LCI, _UCI]
User CI level: 95%, controlled by civarlevel()

Model and method
Model: Random effects
Method: DerSimonian–Laird

Your helpful insights are very much appreciated


Thank you!