model_evaluation {measr}

Add model evaluation metrics model objects

Description

Add model evaluation metrics to fitted model objects. These functions are wrappers around other functions that compute the metrics. The benefit of using these wrappers is that the model evaluation metrics are saved as part of the model object so that time-intensive calculations do not need to be repeated. See Details for specifics.

Usage

add_criterion(
  x,
  criterion = c("loo", "waic"),
  overwrite = FALSE,
  save = TRUE,
  ...,
  r_eff = NA
)

add_reliability(x, overwrite = FALSE, save = TRUE)

add_fit(
  x,
  method = c("m2", "ppmc"),
  overwrite = FALSE,
  save = TRUE,
  ...,
  ci = 0.9
)

add_respondent_estimates(
  x,
  probs = c(0.025, 0.975),
  overwrite = FALSE,
  save = TRUE
)

Arguments

x

A measrfit object.

criterion

A vector of criteria to calculate and add to the model object.

overwrite

Logical. Indicates whether specified elements that have already been added to the estimated model should be overwritten. Default is FALSE.

save

Logical. Only relevant if a file was specified in the measrfit object passed to x. If TRUE (the default), the model is re-saved to the specified file when new criteria are added to the R object. If FALSE, the new criteria will be added to the R object, but the saved file will not be updated.

...

Additional arguments passed relevant methods. See Details.

r_eff

Vector of relative effective sample size estimates for the likelihood (exp(log_lik)) of each observation. This is related to the relative efficiency of estimating the normalizing term in self-normalizing importance sampling when using posterior draws obtained with MCMC. If MCMC draws are used and r_eff is not provided then the reported PSIS effective sample sizes and Monte Carlo error estimates will be over-optimistic. If the posterior draws are independent then r_eff=1 and can be omitted. The warning message thrown when r_eff is not specified can be disabled by setting r_eff to NA. See the relative_eff() helper functions for computing r_eff.

method

A vector of model fit methods to evaluate and add to the model object.

ci

The confidence interval for the RMSEA, computed from the M2

probs

The percentiles to be computed by the ⁠[stats::quantile()]⁠ function to summarize the posterior distributions of each person parameter. Only relevant if method = "mcmc" was used to estimate the model.

Details

For add_respondent_estimates(), estimated person parameters are added to the ⁠$respondent_estimates⁠ element of the fitted model.

For add_fit(), model and item fit information are added to the ⁠$fit⁠ element of the fitted model. This function wraps fit_m2() to calculate the M₂ statistic (Hansen et al., 2016; Liu et al., 2016) and/or fit_ppmc() to calculate posterior predictive model checks (Park et al., 2015; Sinharay & Almond, 2007; Sinharay et al., 2006; Thompson, 2019), depending on which methods are specified. Additional arguments supplied to ... are passed to fit_ppmc().

For add_criterion(), relative fit criteria are added to the ⁠$criteria⁠ element of the fitted model. This function wraps loo() and/or waic(), depending on which criteria are specified, to calculate the leave-one-out (LOO; Vehtari et al., 2017) and/or widely applicable information criteria (WAIC; Watanabe, 2010) to fitted model objects. Additional arguments supplied to ... are passed to loo::loo.array() or loo::waic.array().

For add_reliability(), reliability information is added to the ⁠$reliability⁠ element of the fitted model. Pattern level reliability is described by Cui et al. (2012). Classification reliability and posterior probability reliability are described by Johnson & Sinharay (2018, 2020), respectively. This function wraps reliability().

Value

A modified measrfit object with the corresponding slot populated with the specified information.

References

Cui, Y., Gierl, M. J., & Chang, H.-H. (2012). Estimating classification consistency and accuracy for cognitive diagnostic assessment. Journal of Educational Measurement, 49(1), 19-38. doi:10.1111/j.1745-3984.2011.00158.x

Hansen, M., Cai, L., Monroe, S., & Li, Z. (2016). Limited-information goodness-of-fit testing of diagnostic classification item response models. British Journal of Mathematical and Statistical Psychology, 69(3), 225-252. doi:10.1111/bmsp.12074

Johnson, M. S., & Sinharay, S. (2018). Measures of agreement to assess attribute-level classification accuracy and consistency for cognitive diagnostic assessments. Journal of Educational Measurement, 55(4), 635-664. doi:10.1111/jedm.12196

Johnson, M. S., & Sinharay, S. (2020). The reliability of the posterior probability of skill attainment in diagnostic classification models. Journal of Educational and Behavioral Statistics, 45(1), 5-31. doi:10.3102/1076998619864550

Liu, Y., Tian, W., & Xin, T. (2016). An application of M₂ statistic to evaluate the fit of cognitive diagnostic models. Journal of Educational and Behavioral Statistics, 41(1), 3-26. doi:10.3102/1076998615621293

Park, J. Y., Johnson, M. S., Lee, Y-S. (2015). Posterior predictive model checks for cognitive diagnostic models. International Journal of Quantitative Research in Education, 2(3-4), 244-264. doi:10.1504/IJQRE.2015.071738

Sinharay, S., & Almond, R. G. (2007). Assessing fit of cognitive diagnostic models. Educational and Psychological Measurement, 67(2), 239-257. doi:10.1177/0013164406292025

Sinharay, S., Johnson, M. S., & Stern, H. S. (2006). Posterior predictive assessment of item response theory models. Applied Psychological Measurement, 30(4), 298-321. doi:10.1177/0146621605285517

Thompson, W. J. (2019). Bayesian psychometrics for diagnostic assessments: A proof of concept (Research Report No. 19-01). University of Kansas; Accessible Teaching, Learning, and Assessment Systems. doi:10.35542/osf.io/jzqs8

Vehtari, A., Gelman, A., & Gabry, J. (2017). Practical Bayesian model evaluation using leave-one-out cross-validation and WAIC. Statistics and Computing, 27(5), 1413-1432. doi:10.1007/s11222-016-9696-4

Watanabe, S. (2010). Asymptotic equivalence of Bayes cross validation and widely applicable information criterion in singular learning theory. Journal of Machine Learning Research, 11(116), 3571-3594. https://jmlr.org/papers/v11/watanabe10a.html

Examples

cmds_mdm_dina <- measr_dcm(
  data = mdm_data, missing = NA, qmatrix = mdm_qmatrix,
  resp_id = "respondent", item_id = "item", type = "dina",
  method = "optim", seed = 63277, backend = "rstan",
  prior = c(prior(beta(5, 17), class = "slip"),
            prior(beta(5, 17), class = "guess"))
)

cmds_mdm_dina <- add_reliability(cmds_mdm_dina)
cmds_mdm_dina <- add_fit(cmds_mdm_dina, method = "m2")
cmds_mdm_dina <- add_respondent_estimates(cmds_mdm_dina)