Differences

This shows you the differences between two versions of the page.

--- todo [2021/06/09 14:35] – [Miscellaneous Features and Things To Do] Wolfgang Viechtbauer
+++ todo [2021/11/08 16:01] – Wolfgang Viechtbauer
@@ Line 64: / Line 64: @@
   * <del>The ''rma.mv()'' function should allow for one (or multiple?) //continuous// ''inner'' variables in ''random = ~ inner | outer'' for modeling random slopes of moderator variables. This is especially relevant for meta-analyses examining 'dose-response' relationships (assuming there are studies examining the responses to more than two levels of a dose/exposure variable).</del> (DONE!)
-  * Compute SEs of variance components (and covariances/correlations) in ''rma.mv()''. Not that those SEs are really all that useful. <del>Probably more important: Make ''confint()'' work with ''rma.mv'' objects, to provide profile likelihood CIs.</del> (DONE!).
+  * <del>Compute SEs of variance components (and covariances/correlations) in ''rma.mv()''. Not that those SEs are really all that useful.</del> (DONE!) <del>Probably more important: Make ''confint()'' work with ''rma.mv'' objects, to provide profile likelihood CIs.</del> (DONE!).
   * Also, make ''blup()'' <del>and ''ranef()''</del> work with ''rma.mv'' objects (and ''rma.glmm'' objects?).
@@ Line 74: / Line 74: @@
   * <del>Maybe add a ''ranef()'' function (with alias ''random.effects()'' as in the [[http://cran.r-project.org/web/packages/nlme/index.html|nlme]] package) to just extract the predicted random effects (i.e., free of the fixed effects).</del> (DONE!)
-  * The ''rma.glmm()'' function currently uses ±1/2 coding for the random effects when ''model="UM.FS"'' (i.e., an unconditional model with fixed study effects). Maybe provide an option that allows for switching to 0/1 coding?
+  * <del>The ''rma.glmm()'' function currently uses ±1/2 coding for the random effects when ''model="UM.FS"'' (i.e., an unconditional model with fixed study effects). Maybe provide an option that allows for switching to 0/1 coding?</del> (DONE!)
   * <del>Add the difference, ratio, and odds ratio based on //paired// proportions as outcome measures to ''escalc()''.</del> (DONE!)
@@ Line 99: / Line 99: @@
   * Alternatively, consider writing some of the slower parts directly in C++ (via [[http://cran.r-project.org/web/packages/Rcpp/index.html|Rccp]]; see also [[ http://rcpp.org/]] and [[http://adv-r.had.co.nz/Rcpp.html]]).
-  * Various things are currently computed for ''method="FE"'' as if we are fitting an equal-effects model (e.g., residuals, log likelihood, and so on), when in fact the fixed-effects model is more general. But making the appropriate changes would lead to backwards incompatibility, would just confuse most users, and it's not clear if all of the necessary adjustments can be easily implemented.