Ensemble Weightings and Metric Importance

Coh-Metrix Model 2c

This model used Coh-Metrix scores from 7 min narrative writing samples ("I once had a magic pencil and ...") from 124 students in the spring of Grades 2-5 (Mercer et al., 2019) to predict holistic writing quality on the samples (elo ratings calculated from paired comparisons).

Highly correlated Coh-Metrix metrics (r > |.90|) were excluded during pre-processing (see section on Scoring Model Development for more details).

Mercer, S. H., Keller-Margulis, M. A., Faith, E. L., Reid, E. K., & Ochs, S. (2019). The potential for automated text evaluation to improve the technical adequacy of written expression curriculum-based measurement. Learning Disability Quarterly, 42, 117-128. https://doi.org/10.1177/0731948718803296

Algorithm Weightings in Ensemble

Abbreviations:

• all = ensemble model
• pls = partial least squares regression
• mars = bagged multivariate adaptive regression splines
• gbm = stochastic gradient boosted trees
• svm = support vector machines

The table below presents the linear weightings of each algorithm for the ensemble model.

Intercept	pls	mars	gbm	svm
-10.8192	0.0374	0.2735	0.243	0.5377

Metric Importance in Each Algorithm and Ensemble

Each column sums to 100 (so values can be interpreted as % contribution to the model).

Detailed information on Coh-Metrix abbreviations and indices is available here.

Metric	overall	pls	mars	gbm	svm
DESWC	22.58	5.76	48.45	37.37	3.91
WRDVERB	7.27	1.88	23.22	2.92	1.5
DESWLltd	5.72	1.75	16.17	3.85	1.53
CRFCWOa	4.57	1.9	12.16	1.14	2.44
PCNARp	2.04	2.72	0	3.23	2.49
PCDCz	1.78	1.05	0	3.24	2.07
CRFANPa	1.67	1.23	0	2.21	2.3
LSASS1d	1.67	1.5	0	1.62	2.56
WRDHYPn	1.59	2.38	0	3.29	1.58
SYNSTRUTa	1.59	1.74	0	1.28	2.54
LDMTLD	1.57	1.79	0	2.47	1.95
LSAGN	1.57	1.72	0	1.17	2.55
SMCAUSvp	1.53	0.97	0	1.91	2.17
DESSLd	1.52	0.87	0	1.28	2.45
LSAGNd	1.49	2.29	0	0.11	2.81
WRDFRQmc	1.41	2.04	0	1.45	2.06
DESPL	1.41	3.29	0	0.85	2.25
PCVERBz	1.25	1.73	0	0.62	2.13
SYNMEDpos	1.24	1.78	0	0.72	2.08
LSASSp	1.22	1.97	0	0.13	2.28
SMCAUSv	1.16	0.91	0	1.16	1.78
CRFCWO1d	1.15	1.54	0	0.21	2.14
SMCAUSwn	1.09	2.26	0	0.94	1.63
CNCTempx	1.08	0.26	0	0.57	1.92
WRDHYPv	1.06	2.37	0	1.41	1.36
SMCAUSlsa	1.03	1.59	0	1.07	1.5
WRDNOUN	1	2.13	0	1.68	1.11
PCDCp	1	1.81	0	0.23	1.8
PCVERBp	0.93	1.09	0	0.05	1.79
PCTEMPp	0.92	1.68	0	0.51	1.52
LDTTRc	0.9	1.26	0	1.34	1.14
WRDPRP3s	0.87	1.36	0	1.67	0.92
DESWLlt	0.85	2.05	0	1.12	1.07
CNCTemp	0.85	0.85	0	0.9	1.27
RDL2	0.84	2.17	0	0.54	1.31
DRPP	0.83	1.78	0	1.37	0.94
PCCNCz	0.8	1.87	0	0.31	1.35
DRNP	0.8	1.62	0	0.79	1.16
LDTTRa	0.79	2.31	0	0.24	1.34
WRDAOAc	0.78	0.69	0	0.78	1.17
RDFKGL	0.73	2.06	0	0.13	1.27
SYNNP	0.68	0.65	0	0.24	1.23
CNCPos	0.68	0.82	0	0.67	1.03
CNCCaus	0.67	0.73	0	0.63	1.02
WRDADV	0.67	1.9	0	0.66	0.93
PCSYNz	0.66	1.76	0	0.33	1.07
WRDPRO	0.64	0.73	0	0.9	0.84
CNCLogic	0.64	0.41	0	0.71	0.95
PCCNCp	0.64	1.15	0	0	1.22
DRVP	0.63	0.79	0	0.32	1.08
WRDADJ	0.62	1.12	0	0.38	1
SMINTEp	0.62	1.06	0	0.53	0.95
DRNEG	0.6	0.73	0	0.06	1.13
WRDPOLc	0.6	0.89	0	0.61	0.88
WRDHYPnv	0.58	0.02	0	0.17	1.09
WRDFRQa	0.58	0.36	0	0.35	0.99
SYNLE	0.57	0.09	0	0.62	0.87
SMCAUSr	0.56	0.05	0	0.3	1
DRAP	0.55	1.23	0	0.32	0.89
PCREFz	0.53	1.2	0	0.46	0.78
DESWLsy	0.5	0.85	0	0.43	0.76
WRDMEAc	0.49	0.68	0	0.7	0.63
PCSYNp	0.48	1.45	0	0.02	0.86
CNCADC	0.46	1.61	0	0.41	0.63
WRDFRQc	0.42	0.3	0	0.83	0.45
WRDCNCc	0.39	1.3	0	0.39	0.53
DESWLsyd	0.39	0.32	0	0.33	0.63
SMINTEr	0.26	0.37	0	0.11	0.44
PCCONNz	0.23	0.53	0	0.23	0.32
PCREFp	0.23	0.32	0	0.01	0.44
WRDFAMc	0.1	0	0	0.24	0.09
CRFNO1	0.09	0.98	0	0.1	0.08
PCCONNp	0.08	1.18	0	0.04	0.06
WRDPRP3p	0.02	0.44	0	0.03	0