You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
In R/methods-stats-factanal.r, the definition of recover_conference.factanal() implies that half of the inertia is distributed to the rows and half to the columns, but doing an eigendecomposition of a covariance matrix as factanal() does, we see that the full inertia is distributed to both rows and columns.
The text was updated successfully, but these errors were encountered:
As a reminder to myself, at least two things must be checked here: first, that the internally handled "inertia" is indeed the inertia of the data matrix, not its square (the diagonal matrix of the eigenvalue decomposition); and second, that the conference correctly reflects how it is distributed in the recovered matrix factors.
In R/methods-stats-factanal.r, the definition of recover_conference.factanal() implies that half of the inertia is distributed to the rows and half to the columns, but doing an eigendecomposition of a covariance matrix as factanal() does, we see that the full inertia is distributed to both rows and columns.
The text was updated successfully, but these errors were encountered: