summary.dpm() now reports CFI/TLI (and the associated baseline χ²/df) using
the same baseline specification as Stata's xtdpdml. The original lavaan
values remain available as cfi.lavaan/tli.lavaan in the returned fit
measures.parallel::detectCores() returns NA, preventing lavaan option parsing
failures during examples/checks.partial.pre argument to dpm() to use the strategy
proposed by
Allison (2022).lavaan does not allow non-syntactic variable
names.pre() tag surrounding
it. Do not put the entire interaction inside pre().weights argument.y.free = TRUE.dpm() now has the x.free argument, which can allow predictor variables
to have coefficients that vary by wave.dpm objects now have a glance() method for the broom package.x.free or
y.free are used with dpm().This release contains several important updates and a 1.1.x release is likely to be the one submitted to CRAN.
y.free is TRUE.y.lag = 0)y ~ scale(x)).dpm models with update.dpm objects now a have a tidy method via the broom package.As a side note, there is now a testing suite in place to check models
for accuracy/consistency with xtdpdml. That doesn't mean there will be no
bugs, but it should help prevent any regressions.
This is a major release with several breaking changes compared to the initial development release.
Most noticeably, the name of the package has been changed from clfe to dpm.
This was done to better reflect the scope of the package --- the
cross-lagged fixed effects specification (CFLE) is a special case of the
larger group of dynamic panel model (DPM) specifications availed to users of the
package.
Accordingly, what was once the clfe function is now called dpm.
Internally, the dpm class is now an S4 object that contains the lavaan
class. This means that any method implemented for lavaan objects that isn't
explicitly defined by this package should simply treat dpm objects as if they
were lavaan objects.
The summary method now has more options and is more similar to lavaan's
summary in that regard. Of course, the summary output is much cleaner and
more succinct.
The following arguments have been added to dpm():
y.lag: Equivalent to xtdpdml's ylag. Specify which lags of the DV to
use.y.free: Equivalent to xtdpdml's yfree. Allow stability coefficients for
the lagged DV to vary over time.fixed.effects: Comparable to xtdpdml's re. Use fixed effects (TRUE)
or random effects (FALSE) specification.alpha.free: Equivalent to xtdpdml's alphafree. Allow the fixed effects
to vary over time.NEWS.md file to track changes to the package.The panelr package was added as a
dependency, which has some downstream effects.
panel_data function now belongs to panelr.panel_data object type now inherits from tibble, which has pluses
and minuses but I believe more upside than downside.WageData example data is now part of the panelr package.