[R-pkgs] ez version 3.0

Mike Lawrence Mike.Lawrence at dal.ca
Tue Feb 8 17:37:52 CET 2011

Hi folks,

I'd like to announce the release of version 3.0 of the "ez" package.
This package was developed to aid those that are new to statistical
programming. Over the course of several years of helping colleagues
and students learn R, I observed that folks are often initially turned
off R because they have difficulty obtaining SPSS-like results quickly
(SPSS is the dominant environment in my field, psychology). ez
attempts to fill this gap, providing quick and easy analysis and
graphics for common experimental designs. By easing the early portions
of the R learning curve, ez hopes to promote the spread of R as a
means of open source and reproducible analysis.

ez may also be of interest to more advanced users as it includes the
"ezMixed()" function, which automates the assessment of fixed effects
in a mixed effects modelling context, and the "ezPredict()" function,
which obtains predictions for the fixed effects from a mixed effects

****Installing ez****

Version 3.0 of ez requires that you have R 2.12.1 installed, which is
the latest version of R as of this notice. If you have an older
version of R you will need to update R by installing the latest
version (from http://cran.r-project.org/) before installing ez.

Windows and linux users should be able to install the latest version
by running the command:
install.packages( 'ez' )

Mac users should be able to install the latest version by running the commands:
install.packages( c( 'car' , 'reshape2' , 'plyr' , 'ggplot2' ,
'stringr' , 'lme4' , 'Matrix' ) )
install.packages( 'ez' , type='source' , dependencies=F )

Once installed, running the following commands will load ez and bring
up its help page that links to descriptions of all ez's functions:
library( ez )

****Big changes in version 3.0****

- A big rework of "ezANOVA()" to permit more flexibility, including
more nuanced handling of numeric predictor variables, specification of
sums-of-squares types when data is imbalanced, and an option to
compute/return an aov object representing the requested ANOVA for
follow-up contrast analysis. (The latter two features follow from the
discussion at http://stats.stackexchange.com/questions/6208/should-i-include-an-argument-to-request-type-iii-sums-of-squares-in-ezanova)

- An important bugfix for "ezMixed()", which previously permitted
specification of multiple random effects but silently ignored all but
the last!

- A big rework of "ezMixed()", completely changing the output
(including removal of p-values following advice of Pinero & Bates,
2000, and many on the R-SIG-Mixed-Models mailing list) and providing a
new feature whereby the linearity of numeric fixed effects can be
assessed by comparison to higher polynomial degrees.

****Also new****

As noted above, this version fixes a big bug in "ezMixed()" about
which I wish I could have warned users sooner. To facilitate future
rapid notification of users, I've created a discussion group
(http://groups.google.com/group/ez4r) to which users can/should
subscribe to keep up to date on development news. Additionally, I
created a project page on github
(https://github.com/mike-lawrence/ez/issues) where users can submit
bug reports and feature requests. Finally, I encourage users to rate
or review ez (and any other packages you use) at crantastic.org

****Official CHANGES entry****
- added urls in all documentation pointing to the
bug-report/feature-request site
(https://github.com/mike-lawrence/ez/issues) and the discussion group
- changed reshape dependency to reshape2
-     fixed bug such that if detailed=F and car:Anova fails, the first
line of results from stats:aov is cut off.
-     Added more nuanced treatment of numeric variables
-     Added type argument to specify sums-of-squares type selection (1,2, or 3)
-     Added white.adjust argument to permit heteroscedasticity
adjustment (see ?car::Anova)
-     Added return_aov argument to permit returning a stats:aov object
along with results (this was requested by a user seeking to use the
aov object to compute contrasts)
- ezMixed
-     IMPORTANT: Fixed bug such that only the last specified random
effect was implemented
-     fixed bug such that an error occurred if only one fixed effect
was specified
-     changed output format to a list containing a summary data frame,
a list of formulae, a list of errors, a list of warnings, and
(optionally) a list of fitted models
-     Changed format of summary output including removal of p-values
(on the strong advice of many that the p-values from a likelihood
ratio test of a fixed effect is highly questionable)
-     removed the "return_anovas" argument
-     added nuanced ability to explore non-linear effects via addition
of "fixed_poly" and "fixed_poly_max" arguments
- ezPredict
-     Added ability to handle models fit with "I()"'d variables
-     Added stop error when encountering models with "poly()" and no
supplied prediction data frame
- ezBoot
-     Modified the resampling procedure to ensure that any given
resampling of Subjects has at least 2 unique Subjects.
- ezPerm
-     Modified to return only the data frame containing the permutation test.
- ezResample
-     Added to "include" list for general use
- progress_time
-     Added to "include" list for general use
-     fixed bug that prevented accurate timing
- Added functions:
-     progress_timeCI
-         Computes bootstrapped time-remaining quantile estimates for
long and heterogenous computations (eg. employed in ezMixed)

More information about the R-packages mailing list