Creates a summary object of either a MARK model input or model output which includes number of parameters, deviance, AICc, the beta and real parameter estimates and optionally standard errors, confidence intervals and variance-covariance matrices. If there are several groups in the data, the output is structured by group.
# S3 method for mark
summary(object,...,se=FALSE,vc=FALSE,showall=TRUE,show.fixed=FALSE,brief=FALSE)A list with each of the summarized objects that depends on the argument values. Only the first 4 are given if it is a summary of a model that has not been run.
type of model (e.g., CJS)
user define title if any
descriptive name of fitted model
call to make.mark.model used to construct the model
number of fitted parameters
-2xLog Likelihood value
Number of parameters (always the number of columns in design matrix)
Value of over-dispersion constant if not equal to 1
number of estimated parameters from MARK if different than npar
Small sample corrected AIC using npar; named qAICc if chat not equal to 1
Small sample corrected AIC using npar.unadjusted; prefix of q if chat not equal to 1
dataframe of beta parameters with estimate, se, lcl, ucl
variance-covariance matrix for beta
list of lists, dataframes or matrices depending on value of se and the type of model (triangular versus square PIMS) (see details above)
a MARK model object
additional non-specified argument for S3 generic function
if FALSE the real parameter estimates are output in PIM format (eg. triangular format); if TRUE, they are displayed as a list with se and confidence interval
if TRUE the v-c matrix of the betas is included
if FALSE it only returns the values of each unique parameter value
if FALSE, each fixed value given NA; otherwise the fixed real value is used. If se=TRUE, default for show.fixed=TRUE
if TRUE, does not show real parameter estimates
Jeff Laake
The structure of the summary of the real parameters depends on the type of
model and the value of the argument se and showall. If
se=F then only the estimates of the real parameters are shown and
they are summarized the result element reals in PIM format. The
structure of reals depends on whether the PIMS are upper triangular
("Triang") or a row ("Square" although not really square). For the upper
triangular format, the values are passed back as a list of matrices where
the list is a list of parameter types (eg Phi and p) and within each type is
a list for each group containing the pim as an upper triangular matrix
containing the real parameter estimate. For square matrices, reals
is a list of matrices with a list element for each parameter type, but there
is not a second list layer for groups because in the returned matrix each
group is a row in the matrix of real estimates. If se=TRUE then
estimates, standard error (se), lower and upper confidence limits (lcl, ucl)
and a "Fixed" indicator is passed for each real parameter. If the pims for
the model were simplified to represent the unique real parameters (unique
rows in the design matrix), then it is possible to restict the summary to
only the unique parameters with showall=FALSE. This argument only
has an affect if se=TRUE. If showall=FALSE, reals is
returned as a dataframe of the unique real parameters specified in the
model. This does not mean they will all have unique values and it includes
all "Fixed" real parameters and any real parameters that cannot be
simplified in the case of parameters such as "pent" in POPAN or "Psi" in
"Multistrata" that use the multinomial logit link. Use of
showall=FALSE is of limited use but provided for completeness. In
most cases the default of showall=TRUE will be satisfactory. In this
case, reals is a list of dataframes with a list element for each
parameter type. The dataframe contains the estimate, se,lcl, ucl,fixed and
the associated default design data for that parameter (eg time,age, cohort
etc). The advantage of retrieving the reals in this format is that it is
the same regardless of the model, so it enables model averaging the real
parameters over different models with differing numbers of unique real
parameters.