news:news2016

After fixing some issues introduced by some non-backwards compatible changes in a package that metafor makes use of in a few places, I pushed out version 1.9-9 to CRAN (see here). So this is now the official (i.e., CRAN) release. This version includes many smaller tweaks and improvements, plus a few more noteworthy ones:

- Argument
`knha`

in`rma.uni()`

and argument`tdist`

in`rma.glmm()`

and`rma.mv()`

are now superseded by argument`test`

in all three functions. For backwards compatibility, the`knha`

and`tdist`

arguments still work, but are no longer documented. So, use`test="knha"`

or`test="t"`

instead. - One can now also obtain Cook's distances with
`cooks.distance()`

for`rma.mv`

objects. I will add the possibility to examine not only the influence of individual estimates but of multiple estimates later on (e.g., in the context of a multilevel meta-analysis, one may want to know what the influence is of all estimates within a group/cluster). Also, there are some other diagnostic measures that would be useful to add (especially`rstudent()`

). I hope to get around to this soon. - Added
`ranef()`

for extracting the BLUPs of the random effects only (so like`blup()`

but only the random effects). I get quite a number of requests to make`blup()`

(and hence also`ranef()`

) also work with`rma.mv`

objects, so that is also pretty high on my to-do list. - The
`permutest.rma.uni()`

function gains a`permci`

argument, which can be used to obtain permutation-based CIs of the model coefficients. Note that this is computationally very demanding and may take a long time to complete. - One can now obtain GOSH (i.e., graphical display of study heterogeneity) plots based on Olkin et al. (2012) with the
`gosh()`

function.

The full changelog can be found here.

The development version now carries version number 2.0-0, which will become the next official release once a sufficient number of updates have accumulated. Of course you can always just install the development version if you want to stay on top of those updates.

I've written up some notes on how to speed up model fitting when dealing with large datasets and/or models involving a large number of random effects. See: Speeding Up Model Fitting.

A question that comes up on a regular basis is how one can conduct meta-analyses using the "Hunter and Schmidt method" using the metafor package. A discussion around this has been added to the tips and notes section. See: Hunter and Schmidt Method.

I've received several e-mails recently asking about generalizations of $I^2$ to multilevel and multivariate models. A discussion around this has been added to the tips and notes section. See: $\boldsymbol{I^2}$ for Multilevel and Multivariate Models.

I haven't made any entries here for a while, but development of the metafor package continues (time permitting) on the development version, which eventually will turn into release 1.9-9. Those who would like to play around with the development version already can just install it directly from GitHub with `devtools::install_github("wviechtb/metafor")`

(need to install `devtools`

first of course). You can also track changes and follow the development by going to the GitHub repository for the package.

Besides a lot of smaller tweaks and improvements, some of the more interesting changes are the addition of GOSH plots (you can see an example already here: GOSH Plot), `cooks.distance()`

for `rma.mv`

objects (useful for examining the data for influential cases), and the ability to get permutation-based CIs with `permutest.rma.uni()`

(which is computationally rather demanding).

I am also slowly increasing the code coverage of the automated package tests. At the moment, coverage is just a wee bit shy of 80% (see here for detail).

news/news2016.txt ยท Last modified: 2017/03/19 14:43 (external edit)

Except where otherwise noted, content on this wiki is licensed under the following license: CC Attribution-Noncommercial-Share Alike 4.0 International