Add support for multi-sample models

[sfcheung]

To-Do

• [x] Revise lavaan_rerun() to support multiple-group models.
• [x] Revise est_change() to support multiple-group models.
• [x] Revise est_change_raw() to support multiple-group models.
• [x] Revise fit_measures() to support multiple-group models.
• [x] Revise mahalanobis_rerun() to support multiple-group models.
• [x] Revise mahalanobis_predictors() to support multiple-group models.
• [x] Revise influence_stat() to support multiple-group models.
• [x] Revise fit_measures_approx() to support multiple-group models.
• [x] Revise est_change_approx() to support multiple-group models.
• [x] Revise est_change_raw_approx() to support multiple-group models.
• [x] Revise implied_scores() to support multiple-group models.
• [x] Revise influence_stat() to support multiple-group models.
• [x] Update parameter selection for multiple-group models. <- Already supported
• [x] Update all documentation.

Status

Working on the branch multi_sample.

Note

For a multi-sample models, lavaan stores the data as (1) a list of matrix, and (2) a list of case.idx matching the cases to the original dataset.

Possible approach:

• Reconstruct the full dataset from the list of matrix and the list of case.idx.
• In gen_fct_use_lavaan, reconstruct the full dataset and proceed as usual using case index in the full sample.
• In lavaan_rerun, only lines on case ids need to be revised.

Most other functions should work fine whether the model is multi-sample or not.

Mahalanobis distance is computed using within-group means and covariance matrices.

[sfcheung]

78