Movatterモバイル変換

NEWS	R Documentation

mclogit News

It is now possible to estimate models with non-nested(e.g. crossed) random effects. Such models can be specifiedby providing a list of formulas asrandom= argument to themclogit() ormblogit() function.
The left-hand side of conditional logit models can nowmore conveniently specified using the vertical-bar (|)operator.
It is now possible to choose between differentoptimizers to be used in the inner iterations of the MQL/PQLestimator: One can choose betweennlm(),nlminb(),ucminf(), and most techniques provided byoptim().
Withrebase() the baseline category of a modelcan be changed without the need of refitting the model.
mblogit() andmclogit() now have agroups= argument that allows to estimated overdispersion (across groups).
mblogit() andmclogit() now also have anoffset= argument that to add an offset to the model(i.e. a covariate with coeffecient fixed to unity). Formblogit()the offset can be a matrix with a column for each logit equation.
It is now possible to pass a string tomblogit() asdispersion= argument that specifies the method of estimatingoverdispersion.

Singular initial covariance matrices no longer cause errors.
A warning about unreliable results is issued ifanova() is applied to models with random effects.
Estimating of overdispersion with group data now works.

mclogit() andmblogit() check whether the listpassed ascontrol is complete i.e. contains all the relevantnamed arguments.
Aranef() method is provided for objects createdbymclogit() ormblogit().

It is now possible to use the MQL estimation technique as analternative to PQL.
As an alternative to extending a logit model with randomeffects, it is now possible to add an over-dispersion parameter tothe model.
In addition to approximate the ML estimator, MQL and PQLhave a variant that approximates the REML estimator.
There is now asimulate() method for objects returnedbymblogit() ormclogit() (but only for thosewithout random effects).
Predictions from random-effects models estimated using thePQL technique now are now conditional on the random effects(unless requested otherwise).

mclogit() now handles empty responses (i.e. counts thatsum to zero) correclty.
mclogit() now flags non-numeric response vectors as anerror.
predict() now handles scaled independent variables correcty.

summary() shows the number of groups per randomeffects level (if present).
mclogit() andmblogit() with random effects now work withformula=-argumements passed in variables.

The algorithm for fitting random-effects models tended tostop prematurely returning the starting values obtained using amethods of moments. It has been completely refactored and proceedssimilar to the PQL algorithm in Professor Brian Ripley's MASSpackage:
Now an inner step, in which a linear mixed model is fitted to aworking dependent variable is nested into outer step iterations,in which the the working dependent variable isupdated.
Also, the PQL algorithm no longer builds on starting values from ano-random-effects model, because surprisingly this makes thealgorithm more stable and not less.
As a consequence, the algorithm does a much better job at avoidingdivergence or running into numerical difficulties.
The PQL estimator for random-effects model uses a (inverse)Cholesky factor parametrisation, which makes sure thatrandom-effects (Co-)Variance matrices are always positive(semi-)definite.

mclogit now complains about (i.e. throws an errorexception) when the random effects structure cannot be estimated,e.g. because random effects are constant within choice setsand therefore drop out by the formation of conditional logits.

mblogit now handles responses with only two columns.
mblogit now can deal with matrix responses that haverows that sum to zero.
mclogit andmblogit now return a componentnamed "df.residual" instead of "residual.df".

mclogit now supports conditional logit models withrandom slopes.
mblogit now supports multinomial baseline models withrandom intercept and random slopes.

predict methods of objects created bymclogitandmblogit are better in handling missing data.

Newnobs andextractAIC methods formclogit objects, so thatdrop1.default should work with these.
New functionmblogit to fit multinomial baseline logit models.
mclogitmclogit.fit now allow user-provided starting values.

mclogit.fit andmclogit.fit.rePQL are exported to enable their use by other packages.