`R/parseOutput.R`

`extractModelSummaries.Rd`

Parses a group of Mplus model output files (.out extension) for model fit statistics.
At this time, the details extracted are fixed and include: ```
Filename, InputInstructions, Title, Estimator,
LL, BIC, aBIC, AIC, AICC, Parameters, Observations, CFI, TLI, RMSEA_Estimate, RMSEA_90CI_LB, RMSEA_90CI_UB,
RMSEA_pLT05, ChiSqM_Value, ChiSqM_DF, ChiSq_PValue, BLRT_KM1LL, BLRT_PValue, BLRT_Numdraws)
```

. The
infrastructure is in place to allow for user-specified selection of summary statistics in future versions.

extractModelSummaries(target = getwd(), recursive = FALSE, filefilter)

target | the directory containing Mplus output files (.out) to parse OR the single output file to be parsed. Defaults to the current working directory. Example: "C:/Users/Michael/Mplus Runs" |
---|---|

recursive | optional. If |

filefilter | a Perl regular expression (PCRE-compatible) specifying particular
output files to be parsed within |

Returns a `data.frame`

containing model fit statistics for all output files within `directory`

.
The `data.frame`

contains some of the following variables (depends on model type):

Title for the model, specified by the TITLE: command

Filename of the output file

Estimator used for the model (e.g., ML, MLR, WLSMV, etc.)

Log-likelihood of the model

Bayesian Information Criterion

Sample-Size-Adjusted BIC (Sclove, 1987)

Akaike's Information Criterion

Corrected AIC, based on Sugiura (1978) and recommended by Burnham & Anderson (2002)

Deviance Information Criterion. Available in ESTIMATOR=BAYES output.

Number of parameters estimated by the model

Estimated number of parameters in Bayesian output

The number of observations for the model (does not suppport multiple-groups analysis at this time)

Confirmatory Fit Index

Tucker-Lewis Index

Point estimate of root mean squared error of approximation

Lower bound of the 90% Confidence Interval around the RMSEA estimate.

Upper bound of the 90% Confidence Interval around the RMSEA estimate.

Probability that the RMSEA estimate falls below .05, indicating good fit.

Model chi-squared value

Model chi-squared degrees of freedom

Model chi-squared p value

H0 Scaling Correction Factor

Lower bound of 95% confidence interval for the difference between observed and replicated chi-square values

Upper bound of 95% confidence interval for the difference between observed and replicated chi-square values

Posterior predictive p-value

Prior Posterior Predictive P-Value

Number of requested bootstrap draws for TECH14.

Log-likelihood of the K-1 model (one less class) for the Bootstrapped Likelihood Ratio Test (TECH14).

Two times the log-likelihood difference of the models with K and K-1 classes (TECH14).

Difference in the number of parameters for models with K and K-1 classes (TECH14).

P-value of the Bootstrapped Likelihood Ratio Test (TECH14) testing whether the K class model is significantly better than K-1

The number of successful bootstrapped samples used in the Bootstrapped Likelihood Ratio Test

Standardized root mean square residual

For TYPE=TWOLEVEL output, standardized root mean square residual for between level

For TYPE=TWOLEVEL output, standardized root mean square residual for within level

Weighted root mean square residual

Baseline (unstructured) chi-squared value

Baseline (unstructured) chi-squared degrees of freedom

Baseline (unstructured) chi-squared p value

For TYPE=EFA output, the number of factors

TECH11: Number of initial stage random starts for k-1 model

TECH11: Number of final stage optimizations for k-1 model

TECH11: Log-likelihood of the K-1 model used for the Vuong-Lo-Mendell-Rubin LRT

TECH11: 2 * Log-likelihood Difference of K-class vs. K-1-class model for the Vuong-Lo-Mendell-Rubin LRT

TECH11: Difference in number of parameters between K-class and K-1-class model for the Vuong-Lo-Mendell-Rubin LRT

TECH11: Vuong-Lo-Mendell-Rubin LRT mean

TECH11: Vuong-Lo-Mendell-Rubin LRT standard deviation

TECH11: Vuong-Lo-Mendell-Rubin LRT p-value

TECH11: Lo-Mendell-Rubin Adjusted LRT value

TECH11: Lo-Mendell-Rubin Adjusted LRT p-value

Michael Hallquist

if (FALSE) { allExamples <- extractModelSummaries( "C:/Program Files/Mplus/Mplus Examples/User's Guide Examples") }