Dear Stefanou,
Adding a quick note that if you're comfortable with changing
"teaching_level:time" to "teaching_level*time", then the 7th
coefficient from the top in the new output tests the hypothesis you
mentioned.
res2 <- rma.mv(gi ~ 0 + teaching_level*time, vi, random = ~1|study/obs,
data = dd)
Kind regards,
Reza
On Sun, Jan 30, 2022 at 9:18 AM Viechtbauer, Wolfgang (SP)
<wolfgang.viechtbauer at maastrichtuniversity.nl> wrote:
Start by reading: https://wviechtb.github.io/metafor/reference/anova.rma.html https://wviechtb.github.io/metafor/reference/predict.rma.html And lots of examples here: https://www.metafor-project.org/doku.php/tips:testing_factors_lincoms https://www.metafor-project.org/doku.php/tips:multiple_factors_interactions And here: https://www.metafor-project.org/doku.php/tips:models_with_or_without_intercept Best, Wolfgang
-----Original Message-----
From: Stefanou Revesz [mailto:stefanourevesz at gmail.com]
Sent: Sunday, 30 January, 2022 5:18
To: Viechtbauer, Wolfgang (SP)
Cc: R meta
Subject: Re: Difference of difference
Thank you so much, Wolfgang. I must admit that I failed at achieving
this on my own. In my model below, I want to test:
(coef 4 - coef 1) - (coef 5 - coef 2)
Would that be possible either in anova.rma() or in clubSandwich::Wald_test()?
dd <- read.csv("https://raw.githubusercontent.com/fpqq/w/main/1.csv")
dd$obs <- 1:nrow(dd)
res <- rma.mv(gi ~ 0 + teaching_level:time, vi, random = ~1|study/obs,
data = dd)
Many thanks,
Stefanou
On Fri, Jan 28, 2022 at 9:54 AM Viechtbauer, Wolfgang (SP)
<wolfgang.viechtbauer at maastrichtuniversity.nl> wrote:
Hi Stefanou, It's certainly something you can do. anova() can provide a test of such a contrast and predict() will give you the
CI.
Best, Wolfgang
-----Original Message----- From: Stefanou Revesz [mailto:stefanourevesz at gmail.com] Sent: Thursday, 27 January, 2022 22:37 To: R meta Cc: Viechtbauer, Wolfgang (SP) Subject: Difference of difference Dear Wolfang and the list Members, I've done a post-hoc after my 3-level model. Below is part of the results. At Baseline, elementary was larger than the mixed group by .251 (ignore pval). At post-test1, elementary was larger than the mixed group by .467 (ignore
pval).
Shall we again subtract the baseline difference from the post-test1 difference to better understand the difference between mixed and elementary groups at post-test1? If yes, can we then conduct a test on that difference and get CIs for it? Diff SE tval Df pval mixed Baseline - elementary Baseline -0.251 0.277 -0.905 55 0.984 mixed Post-test1 - elementary Post-test1 -0.467 0.251 -1.861 55 0.582 Thank you so much, Stefanou
_______________________________________________ R-sig-meta-analysis mailing list R-sig-meta-analysis at r-project.org https://stat.ethz.ch/mailman/listinfo/r-sig-meta-analysis