Document Model 3 training study

DESCRIPTION

@@ -10,9 +10,10 @@ Description: Imports variables from 'ReaderBench' (Dascalu et al.,
     2018)<doi:10.1007/978-3-319-66610-5_48>, 'Coh-Metrix' (McNamara et
     al., 2014)<doi:10.1017/CBO9780511894664>, and/or 'GAMET' (Crossley et
     al., 2019) <doi:10.17239/jowr-2019.11.02.01> output files; downloads
-    predictive scoring models described in Mercer & Cannon
-    (2022)<doi:10.31244/jero.2022.01.03> and Mercer et
-    al.(2021)<doi:10.1177/0829573520987753>; and generates predicted
+    predictive scoring models documented in Keller-Margulis et al.
+    (2026)<doi:10.35542/osf.io/36dn5_v1>, Mercer & Cannon
+    (2022)<doi:10.31244/jero.2022.01.03>, and Mercer et
+    al. (2021)<doi:10.1177/0829573520987753>; and generates predicted
     writing quality and curriculum-based measurement (McMaster & Espin,
     2007)<doi:10.1177/00224669070410020301> scores.
 License: MIT + file LICENSE

README.md

@@ -119,6 +119,8 @@ Also see the list of code [contributors](https://github.com/shmercer/writeAlizer
 
 ### Journal Articles
 
+Keller-Margulis, M. A., Mercer, S. H., Matta, M., & Duran, B. (2026). *Identifying optimal indicators of performance using automated written-expression curriculum-based measurement: An examination of genre effects* [Preprint]. EdArXiv. https://osf.io/preprints/edarxiv/36dn5_v1/
+
 Matta, M., Keller-Margulis, M. A., & Mercer, S. H. (2025). Improving written-expression curriculum-based measurement feasibility with automated writing evaluation programs. *School Psychology, 40*(6), 707–717. https://doi.org/10.1037/spq0000691
 
 Matta, M., Mercer, S. H., & Keller-Margulis, M. A. (2023). Implications of bias in automated writing quality scores for fair and equitable assessment decisions. *School Psychology, 38*, 173–181. https://doi.org/10.1037/spq0000517

codemeta.json

@@ -2,7 +2,7 @@
   "@context": "https://doi.org/10.5063/schema/codemeta-2.0",
   "@type": "SoftwareSourceCode",
   "identifier": "writeAlizer",
-  "description": "Imports variables from 'ReaderBench' (Dascalu et al., 2018)<doi:10.1007/978-3-319-66610-5_48>, 'Coh-Metrix' (McNamara et al., 2014)<doi:10.1017/CBO9780511894664>, and/or 'GAMET' (Crossley et al., 2019) <doi:10.17239/jowr-2019.11.02.01> output files; downloads predictive scoring models described in Mercer & Cannon (2022)<doi:10.31244/jero.2022.01.03> and Mercer et al.(2021)<doi:10.1177/0829573520987753>; and generates predicted writing quality and curriculum-based measurement (McMaster & Espin, 2007)<doi:10.1177/00224669070410020301> scores.",
+  "description": "Imports variables from 'ReaderBench' (Dascalu et al., 2018)<doi:10.1007/978-3-319-66610-5_48>, 'Coh-Metrix' (McNamara et al., 2014)<doi:10.1017/CBO9780511894664>, and/or 'GAMET' (Crossley et al., 2019) <doi:10.17239/jowr-2019.11.02.01> output files; downloads predictive scoring models documented in Keller-Margulis et al. (2026)<doi:10.35542/osf.io/36dn5_v1>, Mercer & Cannon (2022)<doi:10.31244/jero.2022.01.03>, and Mercer et al. (2021)<doi:10.1177/0829573520987753>; and generates predicted writing quality and curriculum-based measurement (McMaster & Espin, 2007)<doi:10.1177/00224669070410020301> scores.",
   "name": "writeAlizer: Generate Predicted Writing Quality Scores",
   "relatedLink": ["https://shmercer.github.io/writeAlizer/", "https://CRAN.R-project.org/package=writeAlizer"],
   "codeRepository": "https://github.com/shmercer/writeAlizer/",

vignettes/references.bib

@@ -1,3 +1,12 @@
+@misc{Keller-Margulis2026,
+  author       = {Keller-Margulis, Milena A. and Mercer, Sterett H. and Matta, Michael and Duran, Brenda},
+  title        = {Identifying optimal indicators of performance using automated written-expression curriculum-based measurement: An examination of genre effects},
+  year         = {2026},
+  type         = {Preprint},
+  publisher    = {EdArXiv},
+  url          = {https://osf.io/preprints/edarxiv/36dn5_v1/}
+}
+
 @article{Keller-Margulis2021,
   author  = {Keller-Margulis, Milena A. and Mercer, Sterett H. and Matta, Michael},
   title   = {Validity of automated text evaluation tools for written-expression curriculum-based measurement: A comparison study},

vignettes/scoring-model-development.Rmd

@@ -2053,11 +2053,11 @@ Each column sums to approximately 100, allowing for rounding (values describe re
 
 ReaderBench Model 3, recommended for current use, is an ensemble (formed by averaging predicted quality scores) of three genre-specific models, detailed below.  
 
-The models were trained on ReaderBench scores from 15 min narrative, expository, and persuasive writing samples from students in Grades 2-5 to predict holistic writing quality on the samples (theta scores calculated from paired comparisons).
+The models were trained on ReaderBench scores from 15 min narrative, expository, and persuasive writing samples from students in Grades 3–5 to predict holistic writing quality on the samples (theta scores calculated from paired comparisons) [@Keller-Margulis2026].
 
 Highly correlated ReaderBench metrics (_r_ > |.90|) were excluded during pre-processing (see section on [Scoring Model Development](#scoring-model-development) for more details).
 
-More details on the sample will be provided once peer review is complete on the main study using this model.
+See [@Keller-Margulis2026] for additional details on the study sample, writing tasks, human scoring, and model-development procedure.
 
 ## ReaderBench Model 3narr
 
@@ -3825,11 +3825,11 @@ Each column sums to approximately 100, allowing for rounding (values describe re
 
 Coh-Metrix Model 3, recommended for current use, is an ensemble (formed by averaging predicted quality scores) of three genre-specific models, detailed below.  
 
-The models were trained on Coh-Metrix scores from 15 min narrative, expository, and persuasive writing samples from students in Grades 2-5 to predict holistic writing quality on the samples (theta scores calculated from paired comparisons).
+The models were trained on Coh-Metrix scores from 15 min narrative, expository, and persuasive writing samples from students in Grades 3–5 to predict holistic writing quality on the samples (theta scores calculated from paired comparisons) [@Keller-Margulis2026].
 
 Highly correlated CohMetrix metrics (_r_ > |.90|) were excluded during pre-processing (see section on [Scoring Model Development](#scoring-model-development) for more details).
 
-More details on the sample will be provided once peer review is complete on the main study using this model.
+The training sample, writing tasks, and paired-comparison writing-quality criterion are described in [@Keller-Margulis2026].
 
 ## CohMetrix Model 3narr
 

vignettes/writealizer-getting-started.Rmd

@@ -205,6 +205,8 @@ ReaderBench Model 3 keys also accept an explicit `_v2` suffix; the output column
 | `coh_mod2` | `pred_coh_mod2a` through `pred_coh_mod2c` | `pred_coh_mod2_mean` |
 | `coh_mod1` | `pred_coh_mod1a` through `pred_coh_mod1f` | `pred_coh_mod1_mean` |
 
+The ReaderBench and Coh-Metrix Model 3 families were trained on the same Grades 3–5 sample of 15-minute narrative, expository, and persuasive writing described in [@Keller-Margulis2026].
+
 Model 2 is a simplified version of Model 1. Published Model 1 applications include [@Keller-Margulis2021; @Matta2022].
 
 ### GAMET and the offline demonstration