New version of TripleR: Better standard errors

[Felix Schönbrodt]

Dear PERSOC list members and friends,

Stefan Schmukle and I worked on an update to TripleR. Although only a small jump in version number, this is quite a major update.

• New computation of significance tests in multiple groups.
• Standard errors and p-values of multiple groups are now based by averaging the single-group SE derived by Lashley & Bond (1997). This gives more accurate (and, BTW, typically more narrow) standard errors, especially in the case of very few groups
• PLEASE NOTE: We changed the default behavior. From version 1.5 on, you get the new Lashley & Bond (1997) standard errors by default.
• You can get the old SOREMO-style behavior by setting RR(..., se = "SOREMO")
• Fixed a minor bug in bivariate computations of single-group Lashley-Bond standard errors
• Improved output

For example, you can compare both computations:

data(multiLikingLong)
#manifest bivariate SRM analysis
(RR2m <- RR(liking_a + metaliking_a ~ perceiver.id*target.id|group.id,
    data=multiLikingLong, se="SOREMO"))
(RR2m <- RR(liking_a+metaliking_a ~ perceiver.id*target.id|group.id,
    data=multiLikingLong, se="LashleyBond"))

We recommend to always use the new standard errors.

You can get the current version of TripleR from CRAN:

update.packages("TripleR")

Best,
Felix Schönbrodt & Stefan Schmukle


_______________________________
PD Dr. Felix Schönbrodt
Ludwig-Maximilians-Universität München, Germany

http://www.nicebread.de
http://www.shinyapps.org
http://www.psy.lmu.de/allg2
@nicebread303

`Please cite R packages you use, just as you would cite regular papers!
E.g.: citation('TripleR')`