[R-meta] Adjusting for small sample sizes in metafor

[Viechtbauer Wolfgang (SP)]

Dear Celia,

Regarding your questions/observations:

1) This is correct. When using raw-score standardization, the standardized mean change is computed with (mean1 - mean2) / sd1 (with bias correction) and this is not affected by the correlation; only the sampling variance is affected.

2) The 'vtype' argument is only relevant for measures where it is explicitly mentioned in the help file (see help(escalc)). It is not mentioned in the section on "SMCR", so this argument has no relevance here.

Best,
Wolfgang

>-----Original Message-----
>From: R-sig-meta-analysis [mailto:r-sig-meta-analysis-bounces at r-
>project.org] On Behalf Of Célia Sofia Moreira
>Sent: Wednesday, 07 February, 2018 16:44
>To: r-sig-meta-analysis at r-project.org
>Subject: [R-meta] Adjusting for small sample sizes in metafor
>
>Dear Professor Wolfgang,
>
>I have a basic question regarding effects calculation in a pretest-
>psttest
>control design, using metafor. I'm following your "Morris (2008)" example
>to calculate effect sizes and variances. However, I do not have
>pretest-posttest correlations, and so, I considered a reasonable range of
>correlation values -r-, together with a sensitive analysis. Moreover, I
>have small sample sizes and so I want to adjust sampling variance
>estimates. Concerning the outputs, I obtained that:
>
>1) Changing r only affects the variance estimates, not the effect sizes
>2) The unbiased estimates of the sampling variances (vtype="UB") provides
>the same results. I mean that
>
>datT <- escalc
><http://finzi.psych.upenn.edu/library/metafor/html/escalc.html>(measure="
>SMCR",
>m1i=m_post, m2i=m_pre, sd1i=sd_pre, ni=ni, ri=ri, data
><http://stat.ethz.ch/R-manual/R-devel/library/utils/html/data.html>=datT)
>
>and
>
>datT <- escalc
><http://finzi.psych.upenn.edu/library/metafor/html/escalc.html>(measure="
>SMCR",
>m1i=m_post, m2i=m_pre, sd1i=sd_pre, ni=ni, ri=ri, vtype="UB", data
><http://stat.ethz.ch/R-manual/R-devel/library/utils/html/data.html>=datT)
>
>provide the same sampling variance estimates.
>
>Should I be concerned about these two facts?
>
>Thank you very much,
>celia