`R/modelcomp.R`

, `R/extract.R`

, `R/summary.GDINA.R`

`modelcomp.Rd`

This function evaluates whether the saturated G-DINA model can be replaced by reduced CDMs without significant loss in model data fit for each item using the Wald test, likelihood ratio (LR) test or Lagrange multiplier (LM) test. For Wald test, see de la Torre (2011), de la Torre and Lee (2013), Ma, Iaconangelo and de la Torre (2016) and Ma & de la Torre (2018) for details. For LR test and a two-step LR approximation procedure, see Sorrel, de la Torre, Abad, and Olea (2017), Ma (2017) and Ma & de la Torre (2018). For LM test, which is only applicable for DINA, DINO and ACDM, see Sorrel, Abad, Olea, de la Torre, and Barrada (2017). This function also calculates the dissimilarity between the reduced models and the G-DINA model, which can be viewed as a measure of effect size (Ma, Iaconangelo & de la Torre, 2016).

modelcomp(GDINA.obj = NULL, method = "Wald", items = "all", p.adjust.methods = "holm", models = c("DINA", "DINO", "ACDM", "LLM", "RRUM"), decision.args = list(rule = "simpler", alpha.level = 0.05, adjusted = FALSE), DS = FALSE, Wald.args = list(SE.type = 2, varcov = NULL), LR.args = list(LR.approx = FALSE), LM.args = list(reducedMDINA = NULL, reducedMDINO = NULL, reducedMACDM = NULL, SE.type = 2)) # S3 method for modelcomp extract(object, what = c("stats", "pvalues", "adj.pvalues", "df", "DS", "selected.model"), digits = 4, ...) # S3 method for modelcomp summary(object, ...)

GDINA.obj | An estimated model object of class |
---|---|

method | method for item level model comparison; can be |

items | a vector of items to specify the items for model comparsion |

p.adjust.methods | adjusted p-values for multiple hypothesis tests. This is conducted using |

models | a vector specifying which reduced CDMs are possible reduced CDMs for each item. The default is "DINA","DINO","ACDM","LLM",and "RRUM". |

decision.args | a list of options for determining the most appropriate models including (1) |

DS | whether dissimilarity index should be calculated? |

Wald.args | a list of options for Wald test including (1) |

LR.args | a list of options for LR test including for now only |

LM.args | a list of options for LM test including |

object | object of class |

what | argument for S3 method |

digits | How many decimal places in each number? The default is 4. |

... | additional arguments |

an object of class `modelcomp`

. Elements that can be
extracted using `extract`

method include

- stats
Wald or LR statistics

- pvalues
p-values associated with the test statistics

- adj.pvalues
adjusted p-values

- df
degrees of freedom

- DS
dissimilarity between G-DINA and other CDMs

After the test statistics for each reduced CDM were calculated for each item, the
reduced models with p values less than the pre-specified alpha level were rejected.
If all reduced models were rejected for an item, the G-DINA model was used as the best model;
if at least one reduced model was retained, two diferent rules can be implemented for selecting
the best model specified in argument `decision.args`

:

(1) when `rule="simpler"`

, which is the default,

If (a) the DINA or DINO model was one of the retained models, then the DINA or DINO model with the larger p value was selected as the best model; but if (b) both DINA and DINO were rejected, the reduced model with the largest p value was selected as the best model for this item. Note that when the p-values of several reduced models were greater than 0.05, the DINA and DINO models were preferred over the A-CDM, LLM, and R-RUM because of their simplicity. This procedure is originally proposed by Ma, Iaconangelo, and de la Torre (2016).

(2) When `rule="largestp"`

:

The reduced model with the largest p-values is selected as the most appropriate model.

`extract`

: extract various elements from`modelcomp`

objects`summary`

: print summary information

de la Torre, J., & Lee, Y. S. (2013). Evaluating the wald test for item-level comparison of
saturated and reduced models in cognitive diagnosis. *Journal of Educational Measurement, 50*, 355-373.

Ma, W., Iaconangelo, C., & de la Torre, J. (2016). Model similarity, model selection and attribute classification.
*Applied Psychological Measurement, 40*, 200-217.

Ma, W. (2017). *A Sequential Cognitive Diagnosis Model for Graded Response: Model Development, Q-Matrix Validation,and Model Comparison. Unpublished doctoral dissertation.* New Brunswick, NJ: Rutgers University.

Ma, W. & de la Torre, J. (2018). Category-level model selection for the sequential G-DINA model. *Journal of Educational and Behavorial Statistics*.

Sorrel, M. A., Abad, F. J., Olea, J., de la Torre, J., & Barrada, J. R. (2017). Inferential Item-Fit Evaluation in Cognitive Diagnosis Modeling. *Applied Psychological Measurement, 41,* 614-631.

Sorrel, M. A., de la Torre, J., Abad, F. J., & Olea, J. (2017). Two-Step Likelihood Ratio Test for Item-Level Model Comparison in Cognitive Diagnosis Models. *Methodology, 13*, 39-47.

if (FALSE) { dat <- sim10GDINA$simdat Q <- sim10GDINA$simQ # --- GDINA model ---# fit <- GDINA(dat = dat, Q = Q, model = "GDINA") fit ################### # # Wald test # ################### w <- modelcomp(fit) w # wald statistics extract(w,"stats") #p values extract(w,"pvalues") # selected models extract(w,"selected.model") ########################## # # LR and Two-step LR test # ########################## lr <- modelcomp(fit,method = "LR") lr TwostepLR <- modelcomp(fit,items =c(6:10),method = "LR",LR.args = list(LR.approx = TRUE)) TwostepLR ########################## # # LM test # ########################## dina <- GDINA(dat = dat, Q = Q, model = "DINA") dino <- GDINA(dat = dat, Q = Q, model = "DINO") acdm <- GDINA(dat = dat, Q = Q, model = "ACDM") lm <- modelcomp(method = "LM",LM.args=list(reducedMDINA = dina, reducedMDINO = dino, reducedMACDM = acdm)) lm }