--- title: "A Rasch analysis workflow with easyRasch2" output: rmarkdown::html_vignette bibliography: references.bib link-citations: yes vignette: > %\VignetteIndexEntry{A Rasch analysis workflow with easyRasch2} %\VignetteEngine{knitr::rmarkdown} %\VignetteEncoding{UTF-8} --- ```{=html} ``` ## Overview This vignette walks through a worked Rasch analysis of the nine-item Patient Health Questionnaire (PHQ-9; @kroenke_phq9_2001) using the **easyRasch2** package. The example follows the four psychometric criteria proposed by @christensen_psychometric_2021 for the validation of patient-reported outcome measures (PROMs): 1. **Unidimensionality** --- the items measure a single latent construct. 2. **Local independence** --- after conditioning on the latent trait, item responses are independent of each other. 3. **Ordered response category thresholds** (monotonicity) --- moving up the latent trait increases the probability of higher categories. 4. **Invariance / no DIF** --- item parameters are the same across relevant external groups (e.g. gender). We then move on to two complementary descriptors that are commonly reported alongside the four criteria above: * **Targeting** --- how well person and item locations overlap on the latent continuum. * **Reliability** --- how precisely the scale separates respondents. For a more extensive treatment of Rasch analysis in R, see . For a Bayesian sibling package, see > **NOTE:** all simulation-based functions use a low number of iterations to make this vignette render faster. You should use more iterations for actual analysis work. For most methods, 500-1000 will be useful, except for conditional infit, where 100-400 are optimal, depending on sample size [@johansson_detecting_2025]. ## Data The bundled `phq9` dataset is a 600-respondent random subsample of the PHQ-9 module from the U.S. National Health and Nutrition Examination Survey (NHANES, September 2024 release) with complete responses on all nine items. NHANES microdata are released to the public domain by the U.S. federal government. ``` r library(easyRasch2) data(phq9) items <- phq9[, paste0("q", 1:9)] # 9 item columns, scored 0..3 gender <- phq9$gender # external grouping variables age <- phq9$age # # add item information item_desc <- c( "Little interest or pleasure in doing things", "Feeling down, depressed, or hopeless", "Trouble falling or staying asleep, or sleeping too much", "Feeling tired or having little energy", "Poor appetite or overeating", "Feeling bad about yourself - or that you are a failure or have let yourself or your family down", "Trouble concentrating on things, such as reading the newspaper or watching television", "Moving or speaking so slowly that other people could have noticed?", "Thoughts that you would be better off dead or of hurting yourself in some way" ) item_resp <- c("Not at all","Several days","More than \nhalf the days","Nearly every day") ``` ``` r str(items) #> 'data.frame': 600 obs. of 9 variables: #> $ q1: int 3 0 1 2 3 3 1 3 2 1 ... #> $ q2: int 3 0 2 3 3 3 1 3 2 0 ... #> $ q3: int 3 1 3 0 3 1 0 3 2 0 ... #> $ q4: int 3 1 3 2 3 3 1 3 2 0 ... #> $ q5: int 3 0 3 2 3 2 0 1 2 0 ... #> $ q6: int 3 2 3 2 3 2 2 3 3 0 ... #> $ q7: int 3 3 3 2 3 2 2 3 3 0 ... #> $ q8: int 1 0 2 0 3 3 0 0 1 0 ... #> $ q9: int 3 0 0 2 3 1 2 0 0 0 ... summary(rowSums(items)) #> Min. 1st Qu. Median Mean 3rd Qu. Max. #> 0.00 10.00 16.00 15.41 21.00 27.00 table(gender, useNA = "ifany") #> gender #> Female Male #> 426 143 31 summary(age) #> Min. 1st Qu. Median Mean 3rd Qu. Max. #> 15.00 26.00 33.00 36.06 44.00 85.00 ``` ### Descriptive plots Before fitting any model it is worth eyeballing the response distributions: ``` r hist(rowSums(items), col = "lightblue", main = "") ```

**Figure 1.** *Histogram of ordinal sum scores*

``` r RMplotBar(items, ncol = 2) ```

**Figure 2.** *Faceted bar chart of response distributions*

``` r RMplotTile(items, category_labels = item_resp) ```

**Figure 3.** *Response distribution tile plot*

``` r RMplotStackedbar(items, show_percent = TRUE) ```

**Figure 4.** *Stacked-bar response distribution*

## 1. Unidimensionality `easyRasch2` provides several complementary unidimensionality diagnostics that can be combined for a robust conclusion: - item-level conditional infit MSQ statistics [@muller_item_2020] - item-level item-restscore associations with Goodman-Kruskal's gamma [@kreiner_note_2011] - confirmatory factor analysis (CFA) with WLSMV estimator for ordinal data - principal components analysis (PCA) of the standardised residuals [@chou_checking_2010] - Martin-Löf test with Monte-Carlo p-values [@christensenMonteCarloApproach2007] ### Conditional infit MSQ Conditional item infit mean-square statistics flag items whose response patterns deviate from the Rasch expectation. With `RMitemInfitCutoff()`, per-item highest-density intervals serve as the reference instead rule-of-thumb cutoffs [@johansson_detecting_2025]. ``` r infit_cut <- RMitemInfitCutoff(items, iterations = 100, parallel = FALSE, seed = 3) RMitemInfit(items, cutoff = infit_cut) ``` Table: MSQ values based on conditional estimation (n = 600 complete cases). Cutoff values based on 100 simulation iterations (99.9% HDCI). |Item | Infit MSQ| Infit low| Infit high|Flagged | Relative location| |:----|---------:|---------:|----------:|:-------|-----------------:| |q1 | 0.946| 0.891| 1.125|FALSE | -0.58| |q2 | 0.778| 0.887| 1.157|TRUE | -0.78| |q3 | 1.234| 0.854| 1.199|TRUE | -0.85| |q4 | 0.835| 0.793| 1.156|FALSE | -1.51| |q5 | 1.069| 0.894| 1.119|FALSE | -0.60| |q6 | 0.895| 0.858| 1.090|FALSE | -0.78| |q7 | 0.986| 0.843| 1.170|FALSE | -0.68| |q8 | 1.260| 0.835| 1.198|TRUE | 0.95| |q9 | 1.315| 0.832| 1.136|TRUE | 0.77| You can also get a plot summarizing simulated and observed item infit, using `RMitemInfitCutoffPlot()`. Since conditional infit needs complete data, there is a sibling function that uses multiple imputation with infit that is useful if you have partial missingness in your data - `RMitemInfitMI()` and `RMitemInfitCutoffMI()`. It is important to note that the `RIitemfit()` function uses **conditional** infit, which is both robust to different sample sizes and makes ZSTD unnecessary [@muller_item_2020]. Müller also questions the usefulness of outfit, and my simulation study [@johansson_detecting_2025] reached the same conclusion. Thus, outfit is not reported unless requested. A low item fit value, often referred to as an item being "overfit" to the Rasch model, indicates that responses may be too predictable. This is often the case for items that are very general/broad in scope in relation to the latent variable, for instance asking about feeling depressed in a depression questionnaire. You will often find overfitting items to also have residual correlations (local dependencies) with other items. Overfit may be likened to having a much stronger factor loading than other items in a confirmatory factor analysis or a higher level of discrimination in an Item Response Theory model with two or more parameters. A high item fit value, often referred to as being "underfit" to the Rasch model, can indicate several things. Often underfit is due to multidimensionality or a question that is difficult to interpret and thus has noisy response data. The latter could for instance be caused by a question that asks about two things at the same time, or is ambiguous for other reasons. Next is a visual presentation of conditional item fit across the latent continuum, with respondents split into groups based on their total score. ``` r RMitemICCPlot(items, class_intervals = 5) ```

**Figure 5.** *Conditional ICCs with five class intervals*

### Item-restscore Item-restscore uses Goodman-Kruskal's gamma and shows the expected and observed correlation between an item and a score based on the rest of the items [@kreiner_note_2011]. Similarly, but inverted, to item infit, a lower observed correlation value than expected indicates underfit, that the item may not belong to the dimension. A higher than expected observed value indicates an overfitting and possibly redundant item. Overfitting items will often also show issues with local dependency. Compared to infit, item-restscore more often flags overfit items (based on experience), and less often flags underfit items (based on a simulation study [@johansson_detecting_2025]). ``` r RMitemRestscore(items) ``` |Item | Observed value| Expected value| Difference| Adj. p-value (BH)|p-value sign. | Location| Rel. location| |:----|--------------:|--------------:|----------:|-----------------:|:-------------|--------:|-------------:| |q1 | 0.66| 0.62| 0.04| 0.210| | 0.17| -0.58| |q2 | 0.72| 0.62| 0.10| 0.000|*** | -0.04| -0.78| |q3 | 0.57| 0.63| -0.06| 0.085|. | -0.11| -0.85| |q4 | 0.71| 0.62| 0.09| 0.000|*** | -0.76| -1.51| |q5 | 0.62| 0.62| 0.00| 0.968| | 0.14| -0.60| |q6 | 0.69| 0.63| 0.06| 0.021|* | -0.04| -0.78| |q7 | 0.64| 0.62| 0.02| 0.476| | 0.06| -0.68| |q8 | 0.55| 0.63| -0.08| 0.021|* | 1.70| 0.95| |q9 | 0.59| 0.64| -0.05| 0.151| | 1.51| 0.77| ### CFA-based cutoff for CFI / RMSEA `RMdimCFACutoff()` fits a unidimensional ordinal CFA both to the observed data and to data simulated from a unidimensional PCM, and returns parametric-bootstrap cut-offs for model fit indices SRMR, CFI and RMSEA. Model fit values that exceed the simulated cut-off values are *more extreme than is plausible under a unidimensional data generating process*. ``` r cfa_cut <- RMdimCFACutoff(items, iterations = 100, parallel = FALSE, seed = 2) cfa_cut ``` Table: Partial Credit Model posterior-predictive CFA fit-index check. Observed CFA fit (one-factor, lavaan WLSMV, ordered = TRUE) vs simulated null distribution under PCM unidimensionality. n = 600 complete cases, 9 items, 100 parametric-bootstrap iterations. Cutoffs are one-sided at the 99th percentile of the simulated distribution; an item is flagged when the observed value lies in the worst 1% of the null distribution in the unfavourable direction. |Index | Observed| Cutoff|Direction |Flagged | |:-----|--------:|------:|:----------|:-------| |CFI | 0.9623| 0.9977|< 1st pct |TRUE | |RMSEA | 0.1217| 0.0377|> 99th pct |TRUE | |SRMR | 0.0572| 0.0230|> 99th pct |TRUE | Results can also be plotted using `RMdimCFAPlot(cfa_cut)`. ### Residual PCA After fitting the Rasch model, the residuals should contain no further systematic structure. The *largest eigenvalue* of the residual correlation matrix can be considered the headline diagnostic; a value clearly above the simulation-based cut-off suggest a secondary dimension. However, an eigenvalue below the largest value does not by itself support unidimensionality. ``` r pca_cut <- RMdimResidualPCACutoff(items, iterations = 100, parallel = FALSE, seed = 1) RMdimResidualPCA(items, cutoff = pca_cut) ``` Table: Partial Credit Model (600 complete cases, 9 items). Total observed variance: 54.6% explained by measures, 45.4% unexplained (basis for PCA; n = 600 non-extreme cases). First-contrast cutoff = 1.303 based on 100 simulation iterations (99th percentile). |Component | Eigenvalue| Proportion of variance|Flagged | |:---------|----------:|----------------------:|:-------| |PC1 | 1.652| 0.200|TRUE | |PC2 | 1.453| 0.176|TRUE | |PC3 | 1.220| 0.148|FALSE | |PC4 | 0.988| 0.120|FALSE | |PC5 | 0.930| 0.113|FALSE | Also of interest is the plot of item standardised loadings on the first residual contrast and item locations. This figure can be helpful to identify clusters in data, perhaps related to local dependency and/or multidimensionality. ``` r RMdimResidualPCA(items, output = "loadings") ```

**Figure 6.** *Standardised loadings on the first residual contrast*

## 2. Local independence Local independence (LD) can be assessed with multiple methods. Yen's $Q_3$ statistic [@yen_scaling_1984] is the correlation between person-item standardised residuals for every item pair. Pair-wise $Q_3$ values above the simulation-based cut-off flag LD [@christensen2017]. ``` r q3_cut <- RMlocdepQ3Cutoff(items, iterations = 100, parallel = FALSE, seed = 4) RMlocdepQ3(items, cutoff = q3_cut) ``` Table: Dynamic cut-off: 0.037 (mean Q3 = -0.114 + 0.152). Correlations exceeding the cut-off may indicate local dependence. | |q1 |q2 |q3 |q4 |q5 |q6 |q7 |q8 |q9 |above_cutoff | |:--|:-----|:-----|:-----|:-----|:-----|:-----|:-----|:-----|:--|:------------| |q1 | | | | | | | | | | | |q2 |0.25 | | | | | | | | |* | |q3 |-0.19 |-0.2 | | | | | | | | | |q4 |0 |-0.04 |0.07 | | | | | | |* | |q5 |-0.21 |-0.28 |0.07 |0 | | | | | |* | |q6 |-0.18 |0.04 |-0.2 |-0.16 |-0.13 | | | | |* | |q7 |-0.13 |-0.2 |-0.22 |-0.09 |-0.14 |-0.05 | | | | | |q8 |-0.17 |-0.3 |-0.17 |-0.15 |-0.09 |-0.17 |0.07 | | |* | |q9 |-0.14 |0.06 |-0.23 |-0.28 |-0.24 |0.01 |-0.18 |-0.12 | |* | For a more powerful $Q_3$ test, one can use the simulated cutoffs object to plot the expected range of residual correlations for each item-pair and compare with the observed value. We'll limit the output to the 6 item-pairs that deviate the most. ``` r RMlocdepQ3Plot(simfit = q3_cut, data = items, n_pairs = 6) ```

**Figure 7.** *Observed and expected Q3 residuals*

A second perspective on LD is the *partial gamma* coefficient [@kreinerAnalysisLocalDependence2004;@kreiner_validity_2007] between observed item pairs, conditional on the rest-score. Note that this function evaluates both directions of LD, thus the output is two tables. We'll restrict the output to the 6 item-pairs with largest LD deviations. ``` r RMlocdepGamma(items, n_pairs = 6) ``` Table: Partial gamma LD analysis (n = 600 complete cases). Positive gamma indicates positive local dependence between items. Showing top 6 of 36 pairs by |gamma|. Direction 1: rest score = total - Item2. |Item 1 |Item 2 | Partial gamma| Adj. p-value (BH)|p-value sign. | |:------|:------|-------------:|-----------------:|:-------------| |q1 |q2 | 0.531| 0.000|*** | |q4 |q9 | -0.381| 0.000|*** | |q2 |q9 | 0.332| 0.001|*** | |q2 |q8 | -0.323| 0.001|*** | |q7 |q8 | 0.303| 0.001|*** | |q6 |q9 | 0.287| 0.009|** | Table: Partial gamma LD analysis (n = 600 complete cases). Positive gamma indicates positive local dependence between items. Showing top 6 of 36 pairs by |gamma|. Direction 2: rest score = total - Item1. |Item 1 |Item 2 | Partial gamma| Adj. p-value (BH)|p-value sign. | |:------|:------|-------------:|-----------------:|:-------------| |q2 |q1 | 0.577| 0.000|*** | |q9 |q4 | -0.453| 0.000|*** | |q8 |q2 | -0.415| 0.000|*** | |q4 |q3 | 0.361| 0.000|*** | |q5 |q3 | 0.303| 0.000|*** | |q9 |q2 | 0.291| 0.007|** | You can also get simulation-based thresholds for partial gamma LD, using `RMlocdepGammaCutoff()`, which can be used with `RMlocdepGamma()` and also to plot the results with `RMlocdepGammaPlot()` Item pairs flagged by both $Q_3$ and partial gamma are the strongest candidates for further inspection or possible item revision. ## 3. Ordered response category thresholds For a polytomous item to be measuring as intended, the thresholds separating adjacent response categories should be ordered: the threshold from "Not at all" to "Several days" should sit below the one from "Several days" to "More than half the days", and so on. A classical method to assess item response functions is to plot probability of response curves for each item and response category. ``` r RMitemCatProb(items, category_labels = item_resp) ```

**Figure 8.** *Item Probability Function curves*

`RMitemHierarchy()` plots each item's threshold locations on the latent scale, ordered by overall item difficulty. Disordered thresholds appear as overlapping or reversed segments and are a clear signal that the response categories are not being used in the intended order. ``` r RMitemHierarchy(items, item_labels = item_desc) ```

**Figure 9.** *Item-hierarchy*

## 4. Invariance / no DIF We use two complementary DIF assessments. The Andersen likelihood-ratio test [LRT, @andersen_goodness_1973] partitions the sample by an external variable, refits the model in each subgroup, and compares item locations. The partial gamma approach [@kreiner_validity_2007;@christensen_psychometric_2021] looks for an association between item responses and the external variable *conditional on the rest-score*. Both are run on the *gender* variable here (after dropping respondents with missing gender): ``` r keep <- !is.na(gender) items_g <- items[keep, ] gender_g <- droplevels(gender[keep]) ``` ### Andersen LR-test (eRm) ``` r RMdifLR(items_g, dif_var = gender_g, level = "threshold") ```

**Figure 10.** *Andersen LR-test DIF locations by gender*

The plot shows the item threshold locations estimated in each gender group with the corresponding confidence band. ### Partial-gamma DIF ``` r RMdifGamma(items_g, dif_var = gender_g) ``` Table: Partial gamma DIF analysis (n = 569 complete cases). Positive gamma indicates higher scores in higher DIF group levels. |Item | Partial gamma| SE| Lower CI| Upper CI| Adj. p-value (BH)|p-value sign. | |:----|-------------:|-----:|--------:|--------:|-----------------:|:-------------| |q1 | 0.251| 0.104| 0.046| 0.456| 0.148| | |q2 | 0.411| 0.094| 0.227| 0.595| 0.000|*** | |q3 | 0.064| 0.103| -0.138| 0.266| 1.000| | |q4 | -0.092| 0.114| -0.315| 0.131| 1.000| | |q5 | -0.286| 0.092| -0.467| -0.104| 0.018|* | |q6 | -0.155| 0.105| -0.361| 0.050| 1.000| | |q7 | -0.075| 0.102| -0.275| 0.126| 1.000| | |q8 | -0.112| 0.102| -0.311| 0.088| 1.000| | |q9 | 0.180| 0.100| -0.015| 0.376| 0.638| | For a model-based DIF analysis that can handle *continuous* covariates and interactions (e.g. age × gender), see `?RMdifTree` [@strobl_rasch_2015;@henninger_partial_2025]. ## Targeting A targeting plot summarises how well the item-threshold distribution matches the distribution of person locations on the latent scale --- a Wright-map style display. ``` r RMtargeting(items) ```

**Figure 11.** *Person-item targeting*

## Reliability `RMreliability()` reports four reliability metrics: person separation reliability (PSI); Relative Measurement Uncertainty (RMU) estimate derived from posterior person-location uncertainty using plausible values; Cronbach's alpha; and Empirical reliability (using `mirt::empirical_rxx()`. PSI, alpha and empirical can use bootstrap for confidence intervals. All reliability metrics range from 0 to 1, with higher values indicating better separation/precision. ``` r RMreliability(items, draws = 200, rmu_iter = 20, parallel = FALSE, seed = 5) ``` Table: Reliability for 9 items, n = 600. PSI excludes min/max scoring respondents. |Metric | Estimate| Lower (95% HDCI)| Upper (95% HDCI)|Notes | |:----------------|--------:|----------------:|----------------:|:--------------------------| |Cronbach's alpha | 0.886| NA| NA|no bootstrap | |PSI | 0.847| NA| NA|no bootstrap | |Empirical (WLE) | 0.872| NA| NA|no bootstrap | |RMU (WLE) | 0.882| 0.867| 0.897|200 PVs, 20 RMU iterations | For converting ordinal sum-scores to interval-scaled person-location estimates with associated standard errors, use `RMscoreSE()`. ``` r RMscoreSE(items, output = "ggplot") ```

**Figure 12.** *Sum-score to WLE conversion with 95% CIs*

``` r RMscoreSE(items) ``` Table: Person locations via Warm's WLE (CML item parameters from eRm). | Ordinal sum score| Logit score| Logit std.error| |-----------------:|-----------:|---------------:| | 0| -4.469| 0.682| | 1| -3.234| 0.815| | 2| -2.594| 0.748| | 3| -2.138| 0.661| | 4| -1.779| 0.592| | 5| -1.480| 0.540| | 6| -1.224| 0.502| | 7| -1.000| 0.473| | 8| -0.798| 0.450| | 9| -0.613| 0.433| | 10| -0.440| 0.420| | 11| -0.277| 0.410| | 12| -0.119| 0.403| | 13| 0.034| 0.398| | 14| 0.186| 0.396| | 15| 0.338| 0.396| | 16| 0.491| 0.398| | 17| 0.647| 0.402| | 18| 0.806| 0.408| | 19| 0.970| 0.417| | 20| 1.141| 0.431| | 21| 1.320| 0.450| | 22| 1.512| 0.478| | 23| 1.723| 0.516| | 24| 1.968| 0.565| | 25| 2.276| 0.620| | 26| 2.724| 0.659| | 27| 3.700| 0.567| `RMscoreSE()` also has an option for EAP scores (expected á posteriori). ## Where to next * Each `RM*()` function is documented with its own `?function` reference page including a worked example. * The simulation-based cut-offs used above (`RM*cutoff()`) can be parallelised on multiple CPU cores via the `mirai` package; see the relevant help pages. * For a progress bar on time-consuming simulations, add `verbose = TRUE` to the function call. This should not be used when rendering Quarto/Rmd files. ## References