verbose: Details of the 'verbose' argument of various functions

In spaMM: Mixed-Effect Models, with or without Spatial Random Effects

Details of the verbose argument of various functions

Description

This partially documents the usage of the verbose argument in tow contexts: (1) as an argument of the fitting functions; (2) as an argument of confint.HLfit.

The verbosity of the fitting functions is controlled by the verbose["TRACE"] value. The default is TRACE=FALSE (or 0) which is self-explanatory. TRACE=TRUE (or 1) shows values of outer-estimated parameters (and possibly fixed values of parameters that would be outer-estimated), some cryptic progress bar, and the attained value of the likelihood objective function (see inner-vs-outer for these two types of estimates). See Details for other values.

Tracing the progress of residual-dispersion models is distinctly controlled by verbose["phifit"]: see phi-resid.model.

If the fitted model includes a residual-dispersion model, some tracing output for the latter may be confusingly intermingled with tracing output of the mean-response model. The Details below are valid only for the mean-response model.

confint{., verbose=.} will not affect the verbosity of the fits performed by confint (but will rather affect messages specific to the progress of the confint procedure), unless verbose is a named vector. In particular, verbose["TRACE"] may affect the fit (but this has been implemented only for fit objects inheriting from class pois4mlogit). verbose["confint"] may then be used to control confint. Thus, for pois4mlogit fits,
verbose=FALSE and verbose=c(confint=FALSE) will affect only confint per se, in an equivalent way
verbose=c(TRACE=FALSE) will affect only the fits performed to find the intervals, and
verbose=c(TRACE=FALSE, confint=FALSE) will affect both.
Keep in mind that the verbosity of pois4mlogit fits is also controlled by a progress argument.

Details

0<TRACE<1 only shows the cryptic progress bar.
TRACE=1 was explained above.
TRACE=2 will further show information about the progress of Levenberg-Marquardt-like steps for linear-predictor coefficients.
TRACE=3 will further show information about the progress of distinct Levenberg-Marquardt-like steps random effects given fixed-effect coefficients.
TRACE=4 will further show cryptic information about which matrix computations are requested.
TRACE=5 will further report (through a call to the base trace function) the str(.) of the results of such matrix computations.
TRACE=6 will further pause between iterations of the reweighted least-squares algorithm, allowing a browser session to be called.

(assuming no changes since this was first defined)

spaMM documentation built on Sept. 10, 2026, 1:07 a.m.