[R-meta] comparison between escalc("SMCR") and escalc("SMD")
||||ppo@g@mb@rot@ @end|ng |rom gm@||@com
Mon Jan 9 15:59:54 CET 2023
Thank you Wolfgang also for the references!
Yes, I have realized only after sending the email that you did a
(section "standardized mean change (using raw score standardization)
") because Borenstein suggests transforming the change-score metric
into the raw score metric to include both in the same analysis.
Clearly, the sampling variance needs to take into account the
correlation and this is what is done by "SMCR" regardless the sd of
On Mon, 9 Jan 2023 at 15:49, Viechtbauer, Wolfgang (NP)
<wolfgang.viechtbauer using maastrichtuniversity.nl> wrote:
> Hi Filippo,
> Yes, for SMCR, escalc() only uses sd1i in the denominator. One could in principle pool together sd1i and sd2i, but in either case, the denominator represents the SD at a given time point (for pooling, we assume that the true SD is the same at the two timepoints) and NOT the SD of change scores. So, either way, you would be using 'raw score standardization' (and not 'change score standardization'), which is crucial if you would want to combine such values with SMD-type effects.
> If you want to pool sd1i and sd2i, then there has been some recent work to work out the statistical properties of such a measure:
> Cousineau, D. (2020). Approximating the distribution of Cohen's d_p in within-subject designs. The Quantitative Methods for Psychology, 16(4), 418-421. https://doi.org/10.20982/tqmp.16.4.p418
> Cousineau, D., & Goulet-Pelletier, J.-C. (2021). A study of confidence intervals for Cohen's dp in within-subject designs with new proposals. The Quantitative Methods for Psychology, 17(1), 51-75. https://doi.org/10.20982/tqmp.17.1.p051
> I think the CohensdpLibrary package implements this.
> Note that SMCR (or the version that uses the pooled SD) does require the pre-post correlation (for computing the sampling variance).
> >-----Original Message-----
> >From: R-sig-meta-analysis [mailto:r-sig-meta-analysis-bounces using r-project.org] On
> >Behalf Of Filippo Gambarota
> >Sent: Thursday, 05 January, 2023 13:38
> >To: R meta
> >Subject: [R-meta] comparison between escalc("SMCR") and escalc("SMD")
> >I have a very simple question about the comparison between scale ("SMCR")
> >and escalc("SMD"). Carefully reading Morris and De Shon (2002) the
> >comparison between two effect size is meaningful when the denominator
> >represents the same quantity. Assuming that I have a two-group comparison
> >and pre-post comparison on the same meta-analysis I could compute the
> >standard SMD for the first case and the pre-post difference divided by the
> >pooled sd (ignoring the correlation) in the second case. But this is not
> >what "SMCR" is doing because it uses the first "sd1i" which is the "pre"
> >standard deviation.
> >Is that correct? Should I compute the pooled sd before and then supplying
> >it as "sd1i"?
> >*Filippo Gambarota*
> >PhD Student - University of Padova
> >Department of Developmental and Social Psychology
> >Website: filippogambarota.xyz
> >Research Groups: Colab <http://colab.psy.unipd.it/> Psicostat
PhD Student - University of Padova
Department of Developmental and Social Psychology
Research Groups: Colab Psicostat
More information about the R-sig-meta-analysis