# Meta analysis for two identical studies

**URL:** <https://discourse.mc-stan.org/t/meta-analysis-for-two-identical-studies/7493>\
**Category:** brms\
**Created:** [January 29, 2019, 8:39pm UTC](https://discourse.mc-stan.org/t/meta-analysis-for-two-identical-studies/7493 "2019-01-29T20:39:54Z")\
**Posts on this page:** 4\
**Page:** 1

<div class="post-metadata">

**Author:** ![reb.andrews](https://avatars.discourse-cdn.com/v4/letter/r/b9bd4f/32.png) [@reb.andrews](https://discourse.mc-stan.org/u/reb.andrews)\
**Post date:** [January 29, 2019, 8:39pm UTC](https://discourse.mc-stan.org/t/meta-analysis-for-two-identical-studies/7493/1 "2019-01-29T20:39:54Z")

</div>

I have two identical studies that I am trying to meta-analyze using brms. These are basic regression models with two predictors and their interaction and I’d like to synthesize the evidence for each coefficient when taking into account both studies. I have seen several helpful posts on conducting meta-analyses on brms. However, these often have more than a few studies and I am interested in learning more about what I should do if I only have two studies. Should I just set a very constrained prior on the between study variance? Or since the studies are exact replications, would another option be to just use the “posterior\_samples” command to then paste the posterior samples for each sample together? Wouldn’t this be equivalent to a fixed effects meta-analysis?

---

<div class="post-metadata">

**Author:** ![reb.andrews](https://avatars.discourse-cdn.com/v4/letter/r/b9bd4f/32.png) [@reb.andrews](https://discourse.mc-stan.org/u/reb.andrews)\
**Post date:** [January 30, 2019, 3:26pm UTC](https://discourse.mc-stan.org/t/meta-analysis-for-two-identical-studies/7493/2 "2019-01-30T15:26:17Z")

</div>

As an update, I think I can used this brms syntax to model the results for each coefficient:

**y | se(sqrt(v)) ~ 0 + intercept**

Where y is the regression coefficient for each study and v is the estimated error for each coefficient. However, I’m wondering if this is preferable than the method I mentioned in my previous post?

---

<div class="post-metadata">

**Author:** ![paul.buerkner](https://yyz2.discourse-cdn.com/flex030/user_avatar/discourse.mc-stan.org/paul.buerkner/32/3303_2.png) [@paul.buerkner](https://discourse.mc-stan.org/u/paul.buerkner)\
**Post date:** [January 30, 2019, 3:27pm UTC](https://discourse.mc-stan.org/t/meta-analysis-for-two-identical-studies/7493/3 "2019-01-30T15:27:58Z")

</div>

Do you have the raw data for both studies? If yes, I would simply merge the data sets and fit the models using all data (and perhaps include a study indicator as additional predictor).

---

<div class="post-metadata">

**Author:** ![reb.andrews](https://avatars.discourse-cdn.com/v4/letter/r/b9bd4f/32.png) [@reb.andrews](https://discourse.mc-stan.org/u/reb.andrews)\
**Post date:** [January 30, 2019, 3:31pm UTC](https://discourse.mc-stan.org/t/meta-analysis-for-two-identical-studies/7493/4 "2019-01-30T15:31:00Z")

</div>

Yes, I have the raw data. I did both of the studies and they are completely identical. I will merge the data then and then refit the model that I have already fit on each data set.