Package Description ------------------- The GMX package provides the function `gmx()`, which allows for plotting several variants of the graphical model check: - it extends to an arbitrary number of split groups (which could be selected), - it allows for plotting the beta parameters (as does `eRm::plotGOF()` by default), the threshold parameters (for PCM/RSM applications), or the person parameters of the split groups (RM, PCM, and RSM). - users may select subsets of items or parameters to plot, and, - GMX provides further graphical options currently not available in the `eRm::plotGOF()` function. The package requires the `eRm` package and uses the output of the `eRm::LRtest()` function. By default, it extracts the subgroup parameters and draws the graphical model checks of all pairs of split groups in a diagram matrix. - The `type="cumulative"` option (replacing the former `type="beta"` option) allows for selecting the estimates of the cumulative thresholds ($\hat{\beta}$ (in `eRm`-notation), - the `type="thresholds"` option draws the threshold estimates $\hat{\tau}$ (default as of package version 0.9), and - the `type="perspar"` option draws the person parameter estimates $\hat{\theta}_r$ for all scores $r$. As of version 0.9, GMX also supports the psychotools package. It now provides the `cLRT()` function, the result object of which is processed by `gmx()`. Example Code ------------ Examples use the data set `pcmdat2` of `eRm`: library(eRm) data(pcmdat2) mod = PCM(pcmdat2) grp = rep(1:3, each=100) tst = LRtest(mod, splitcr=grp) library(GMX) gmx(tst) # betas (= default) gmx(tst,type="t") # thresholds gmx(tst,type="p",EQlims=FALSE, # person parameters xlim=c(-3,3),ylim=c(-3,3)) palette(c("darkorange2","deepskyblue1","green3","steelblue")) gmx(tst,col="items",type="t",cex=2) # distinguish items palette(c("steelblue1","orangered2")) gmx(tst,col="thresholds",type="t",cex=2) # distinguish thresholds Example Output -------------- * The default plot (betas, i.e., cumulative threshold estimates for polytomous data) obtained with `gmx(tst)`: ![gmx(tst)][1] * The thresholds plot with `gmx(tst, type="t")`: ![gmx(tst,type="t")][2] Note: The options `type="b"` and `type="t"` will yield identical plots for dichotomous data (RM) but may differ largely for polytomous data (PCM/RSM). * The person parameter plot with `gmx(tst, type="p")`: ![gmx(tst,type="p")][3] * Auto-coloring the items with `gmx(tst,col="items",type="t",cex=2)`: ![gmx(tst,col="items",type="t",cex=2)][4] * Auto-coloring the thresholds with `gmx(tst,col="thresholds",type="t",cex=2)`: ![gmx(tst,col="thresholds",type="t",cex=2)][5] Citation -------- If you use this package in your publication, pls. cite the following article: **Alexandrowicz, R. W. (2022). GMX: Extended Graphical Model Checks. A Versatile Replacement of the plotGOF() Function of eRm. *Psychological Test and Assessment Modeling. 64 (3)*, 215--225**. [Download][6] This article explains technical details and usage of the package. The updates of version 0.9 are described in: **Alexandrowicz, R. W. (2023). Extending GMX: Conditional Likelihood Ratio Test and Extended Graphical Model Checks with psychotools. *Psychological Test and Assessment Modeling. 65 (2)*.** Updates ------- **2023-06-09 -- Version 0.9-1** * Bugfixes and minor changes. * gmx: `col=` option renamed to `dotcol=`. **2023-05-19 -- Version 0.9-0:** * GMX now also supports the psychotools package. **2022-09-29 -- Version 0.8-2:** * Minor correction in help file. **2022-09-14 -- Version 0.8-1:** * New option `parskip=TRUE|FALSE` to override internal graphics settings of `par()` giving the user full control over layout, margins, etc. (default: `parskip=FALSE`, i.e., use internal settings). [1]: https://osf.io/ymc6v/download [2]: https://osf.io/wbrxh/download [3]: https://osf.io/nxa52/download [4]: https://osf.io/emp5k/download [5]: https://osf.io/m9w65/download [6]: https://www.psychologie-aktuell.com/fileadmin/Redaktion/Journale/ptam_2022-3/Alexandrowicz_online.pdf