Fixed issues with nested loop plot when the simulation design isnot fully-factorial (#47, thanks
Fixed wrong calculations when the same column was used intrue and inby or elsewhere (#48, thanks
New options for zip plots: via thezip_ci_colorsargument it is possible to customise the colour of CI lines forcoverage. More details are available on GitHub (https://github.com/ellessenne/rsimsum/pull/55) and inthe documentation.
Updated columns names for confidence intervals in thett dataset;
Updated documentation regarding column names that are not allowedwhen callingsimsum() ormultisimsum().Internals of the package have been reworked accordingly, and moreinformative error messages are provided to the user.
A new performance measure, relative bias, can now be calculatedalong with its Monte Carlo error (#41). More details on the formulae arein the introductory vignette, which has been updatedaccordingly.
Fixed issues withstat(level), which was deprecatedin {ggplot2} 3.4.0 (#44).
Fixed an error in the calculation of the Monte Carlo standarderror for relative % error in ModSE (#45, thanks
Several improvements to package documentation.
This is a minor release, with the following changes:
Updated hex sticker.
Updated maintainer e-mail.
Fixed conflicts withtidy() function from the{broom} package when both {broom} and {rsimsum} are loaded at the sametime. This could lead to an error of this kind:
#> Error: No tidy method recognized for this list.Thanks to Theodosia Salika for reporting this.
The {pkgdown} website with the documentation has been updated touse Bootstrap 5 (and {pkgdown} ≥ 2.0.0). The new site can be found here:https://ellessenne.github.io/rsimsum/
Updated a DOI that was returning HTTP Error 503.
print.summary.simsum() now return (invisibly) a listwith each section of the output, e.g. by performance measure. This isuseful for printing small sections of the output, e.g. usingkable() (thanks
library(rsimsum)s2<-simsum(data = relhaz,estvarname ="theta",true =-0.50,se ="se",methodvar ="model",by =c("baseline","n"))out<-print(summary(s2,stats ="bias"))library(knitr)kable(out[[1]],caption =names(out)[1],align ="r")This is implemented forprint.summary.multisimsum() aswell, with an additional level of nesting (by parameter).
.Rmd files), thanks topower_df was passed to thecontrolargument, it was not used (regression introduced in {rsimsum} 0.9.0).Now fixed, thanks toget_data() is now deprecated in favour oftidy();get_data() still works (and is fullytested), but now throws a warning and will be fully removed some time inthe future.simsum() andmultisimsum() now acceptmultiple column inputs that identify unique methods (see e.g. #24, #30).Internally, this combines the unique values from each column factoriallyusing theinteraction() function; then, methods areanalysed and reported as such. Seevignette("E-custom-inputs", package = "rsimsum") for someexamples.
Two new datasets,MIsim2 andfrailty2,are now bundled withrsimsum to test the new functionalityintroduced above. They correspond toMIsim andfrailty, respectively, with the only difference being thatthe (single) column identifying methods is now split into two distinctcolumns.
Improved printing for simulation studies with ‘non-standard’ wayof passing true values (see e.g. #28 on GitHub);
Fixed a typo in introductory vignette;
Some internal housekeeping.
df has been renamed topower_df, and now affects power calculations only.Newdf argument,simsum andmultisimum now accept a column indatacontaining a number of degrees of freedom that will be used to calculateconfidence intervals for coverage (and bias-eliminated coverage) with tcritical values (instead of normal-theory intervals, the defaultbehaviour). Notably, zip plots behave accordingly when calculating andranking confidence intervals;
Calculations for zip plots are noticeably faster now;
Added a simplekable method for objects of classsimsum,summary.simsum,multisimsum,summary.multisimsum to ease thecreation of LaTeX/HTML/Markdown/reStructuredText tables.
byfactors from being plotted.autoplot methods will now plot the number ofnon-missing point estimates/SEs by default (if thestatargument is not set by the user). The previous default was to plot bias,which might not always be available anymore sincersimsum 0.8.0.Handling more plotting edge cases, for instance when standarderrors or true values are not available;
Improvedmultisimsum example in vignette on custominputs.
Added new argumentzoom toautoplotmethods: it is now possible tozoom on the top x% of a zip plotto improve readability;
Added a new example dataset from a toy simulation study assessingthe robustness of the t-test. See?"tt" for moredetails;
Thetrue argument ofrsimsum andmultisimsum now accepts a string that identifies a columnindata. This is especially useful in settings where thetrue value varies across replications, e.g. when it depends oncharacteristics of the simulated data. Seevignette("E-custom-inputs", package = "rsimsum") for moredetails and examples;
Analogously, theci.limits argument now accepts avector of strings that identifies lower and upper limits forcustom-defined confidence intervals from columns indata.Once again, more details are included invignette("E-custom-inputs", package = "rsimsum");
rsimsum now correctly usesinherits(obj, "someclass") instead ofclass(obj) == "someclass" (#20);
Fixed bugs and errors that appeared when auto-plotting results ofsimulation studies with no methods being compared (#23).
autoplot supports two new visualisations: contour plotsand hexbin plots, for either point estimates or standard errors. Theycan be obtained by selecting the argumenttype = "est_density",type = "se_density",type = "est_hex", ortype = "se_hex".Passing the true value of an estimand (trueargument) is no longer required; iftrue is not passed tosimsum ormultisimsum, bias, coverage, andmean squared error are not computed;
Passing estimated standard errors per replication(se argument) is no longer required; if so, average andmedian variances, model-based standard errors, relative error, coverageprobability, bias-eliminated coverage probability, power are notcomputed.
rsimsum 0.6.1 (average andmedian variances were not printed).Fixed labelling bug in zipper plots (thanks to
Clarified thatsimsum andmultisimsumreport average (or median) estimated variances, not standard errors(thanks to Ian R. White for reporting this).
Implemented fully automated nested loop plots for simulationstudies with several data-generating mechanisms:autoplot(object, type = "nlp");
Addeddata("nlp", package = "rsimsum"), a datasetfrom a simulation study with 150 data-generating. This is particularlyuseful to illustrate nested loop plots;
Added a new vignette on nested loop plots;
Improved ordering of vignettes.
Updated unquoting for compatibility withrlang0.4.0;
Fixed missing details and options in the documentation ofautoplot.multisimsum andautoplot.summary.multisimsum.
ggplot2::label_both for ‘by’ factors (whenincluded).autoplot method formultisimsum andsummary.multisimsumobjects;simsum andmultisimsum objects;autoplot methods pick the value oftrue passed tosimsum,multisimsum when inferring the target value ifstats = (thetamean, thetamedian) andtarget = NULL. In plain English, the true value of theestimand is picked as target value when plotting the mean (or median) ofthe estimated value;pkgdown website, published athttps://ellessenne.github.io/rsimsum/;autoplot caused by premature slicing ofby arguments, where noby arguments wereincluded.Implementedautoplot method forsimsum andsummary.simsum objects; when callingautoplotonsummary.simsum objects, confidence intervals based onMonte Carlo standard errors will be included as well (if sensible).
Supported plot types are:
Several options to customise the behaviour ofautoplot,see?autoplot.simsum and?autoplot.summary.simsum for further details.
Fixed a bug indropbig and related internal functionthat was returning standardised values instead of actual observedvalues.
rsimsum 0.4.0 is a large refactoring ofrsimsum. There are several improvements and breakingchanges, outlined below.
rsimsum is more robust to using factor variables(e.g. asmethodvar orby factor), withordering that will be preserved if defined in the dataset passed tosimsum (ormultisimsum);help(summary.simsum) for more details;simsum fortesting purposes - differences are negligible, and there are somecalculations insimsum that are wrong (already reported).Most differences can be attributed to calculations (and conversions, forcomparison) on different scales.simsum andmultisimsum hasbeen slightly changed, with some arguments being removed and othersbeing moved to acontrol list with several tuningparameters. Please check the updated examples for more details;dropbig is no longer an S3 method forsimsum andmultisimsum objects. Now,dropbig is an exported function that can be used toidentify rows of the inputdata.frame that would be droppedbysimsum (ormultisimsum);simsum(ormultisimsum) whendropbig = TRUE) are nolonger included in the returned object; therefore, the S3 methodmiss has been removed;get_data is no longer an S3 method, but still requiresan object of classsimsum,summary.simsum,multisimsum, orsummary.multisimsum to bepassed as input;rsimsum 0.5.0.zip method has been renamed tozipper() to avoid name collision withutils::zip().ci.limits argument (#6,relhaz dataset bundled withrsimsum.rsimsum 0.3.3 focuses on improving the documentation ofthe package.
Improvements: * Improved printing of confidence intervals for summarystatistics based on Monte Carlo standard errors; * Added adescription argument to eachget_data method,to append a column with a description of each summary statisticsexported; defaults toFALSE; * Improved documentation andintroductory vignette to clarify several points (#3,
New: * Added CITATION file with references to paper in JOSS.
rsimsum 0.3.2 is a small maintenance release: * Mergedpull request #1 fromcheckmate package; * Fixed a bug where automatic labels inbar() andforest() were not selectedproperly.
Bug fixes: *bar(),forest(),lolly(),heat() now appropriately pick adiscrete X (or Y) axis scale for methods (if defined) when the methodvariable is numeric; *simsum() andmultisimsum() coercemethodvar variable tostring format (if specified and not already string); * fixed typos forempirical standard errors in documentation here and there.
Updated code of conduct (CONDUCT.md) and contributingguidelines (CONTRIBUTING.md).
Removed dependency on thetidyverse package (thanks MaraAverick).
Bug fixes: *pattern() now appropriately pick a discretecolour scale for methods (if defined) when the method variable isnumeric.
New plots are supported: *forest(), for forest plots; *bar(), for bar plots.
Changes to existing functionality: * thepar argument oflolly.multisimsum is now not required; if not provided,plots will be faceted by estimand (as well as any otherbyfactor); * updatedVisualising results from rsimsumvignette.
AddedCONTRIBUTING.md andCONDUCT.md.
Internal housekeeping.
Added S3 methods forsimsum andmultisimsumobjects to visualise results: *lolly(), for lolly plots; *zip(), for zip plots; *heat(), for heatplots; *pattern(), for scatter plots of estimates vsSEs.
Added a new vignetteVisualising results from rsimsum tointroduce the above-mentioned plots.
Addedx argument tosimsum andmultisimsum to include original dataset as a slot of thereturned object.
Added amiss function for obtaining basic information onmissingness in simulation results.miss has methodsprint andget_data.
First submission to CRAN.rsimsum can handle:
Summary statistics that can be computed are: bias, empirical standarderror, mean squared error, percentage gain in precision relative to areference method, model-based standard error, coverage, bias-correctedcoverage, and power.
Monte Carlo standard errors for each summary statistic can becomputed as well.