Skip to content

Fix mclust covariance parameter counts for EEE and VVV models - #217

Open
zeper-eng wants to merge 1 commit into
data-edu:masterfrom
zeper-eng:fix-mclust-covariance-parameter-count
Open

zeper-eng wants to merge 1 commit into
data-edu:masterfrom
zeper-eng:fix-mclust-covariance-parameter-count

Conversation

@zeper-eng

Copy link
Copy Markdown

calc_fitindices.Mclust() currently counts off-diagonal covariance parameters as p * (p - 1) for EEE and VVV models.

Because covariance matrices are symmetric, the number of unique off-diagonal covariance parameters is p * (p - 1) / 2.

This inflates the reported parameter count and downstream information criteria for Models 3 and 6.

This change:

  • corrects the EEE and VVV parameter-count formulas
  • adds regression tests verifying that tidyLPA's reported parameter count agrees with the underlying mclust model's df

Testing:

  • targeted regression tests: PASS 2 / FAIL 0
  • full test suite: FAIL 0 / PASS 13
  • Mplus-dependent tests were skipped locally because Mplus is not installed on my device.

This branch has not been deployed

No deployments
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant