Forthcoming fixes - cmu-phil/tetrad GitHub Wiki

Changes for forthcoming version 7.6.10

Interim up-to-date beta versions of the Tetrad launch jar with updates listed below can be downloaded using this link in py-tetrad:

https://github.com/cmu-phil/py-tetrad/blob/main/pytetrad/resources/tetrad-current.jar

To download the beta version, click this link and click the button there with the downward-pointing arrow in the upper right. This will save a file called "tetrad-current.jar"; this will be a launchable Tetrad interface jar. Rename this file if you want to distinguish versions.

New features:

  1. Added recursive adjustment sets to the Paths dialog.

  2. Added O-sets (Henckel et al.) to the Paths dialog.

  3. Added Optimal IDA (Witte et al.) to the IDA editor in the Compare box.

  4. Added knowledge sensitivity to DAG to PAG.

  5. Added a new parameter to the LV method, like FCI, to let the user choose whether they want to exclude the possibility of selection bias. This affects required orientations from knowledge, as well as whether the final rules R4, R6, and R7 are performed.

  6. Added a “Save SEM (lavaan)” export option for SEM IMs, which writes the instantiated SEM to lavaan (.lav) syntax with configurable inclusion of intercepts and residual (co)variances and a choice between fixed parameters and lavaan start() values.

  7. Added support for loading and saving covariance matrices in full square-matrix format (while retaining support for the legacy lower-triangular format).

  8. Added an N-tad Explorer data tool that uses canonical correlation to list rank-deficient tetrads, sextads, octads, and higher-order N-tads in a dataset or model-implied covariance matrix, reporting estimated ranks and p-values in an interactive table.

  9. Added an Adjustment & Total Effects regression tool in the Regression box that uses recursive adjustment to enumerate valid adjustment sets and compute linear-regression total effects, supporting both pairwise X–Y effects and joint interventions (p(Y \mid do(X))).

Bug fixes / Technical improvements:

  1. Graph save/load methods for the PCALG amat.pag and amag.cpdag formats were buggy; this has been fixed.

  2. Update MPDAG references to PDAG throughout the code.

  3. Refactored and enhanced Bayes Estimator likelihood calculation to fix p-values for estimated Bayes models.

  4. Fixed a rare bug in the CPDAG to DAG method, affecting scoring for FGES, among other methods.

  5. Fixed several long-standing issues in the Data box editor: column and row deletion now behave consistently, single-cell edits work reliably, and copy/paste of wide blocks (including more than 30 columns and between different data boxes) preserves alignment between variable names and data and expands the table as needed.