summary.sars: Summarising the results of the model fitting functions

View source: R/class_summary.R

summary.sarsR Documentation

Summarising the results of the model fitting functions

Description

S3 method for class 'sars'. summary.sars creates summary statistics for objects of class 'sars'. The exact summary statistics computed depends on the 'Type' attribute (e.g. 'multi') of the 'sars' object. The summary method generates more useful information for the user than the standard model fitting functions. Another S3 method (print.summary.sars; not documented) is used to print the output.

Usage

## S3 method for class 'sars'
summary(object, ...)

Arguments

object

An object of class 'sars'.

...

Further arguments.

Value

The summary.sars function returns an object of class "summary.sars". A print function is used to obtain and print a summary of the model fit results.

For a 'sars' object of Type 'fit', a list with 16 elements is returned that contains useful information from the model fit, including the model parameter table (with t-values, p-values and confidence intervals), model fit statistics (e.g. R2, AIC), the observed shape of the model and whether or not the fit is asymptotic, and the results of any additional model checks undertaken (e.g. normality of the residuals).

For a 'sars' object of Type 'multi', a list with 5 elements is returned: (i) a vector of the names of the models that were successfully fitted and passed any additional checks, (ii) a character string containing the name of the criterion used to rank models, (iii) a data frame of the ranked models, (iv) a vector of the names of any models that were not fitted or did not pass any additional checks, and (v) a logical vector specifying whether the optim convergence code for each model that passed all the checks is zero. In regards to (iii; Model_table), the dataframe contains the fit summaries for each successfully fitted model (including the value of the model criterion used to compare models, the R2 and adjusted R2, and the observed shape of the fit); the models are ranked in decreasing order of information criterion weight.

For a 'sars' object of Type 'lin_pow', a list with up to 7 elements is returned: (i) the model fit output from the lm function, (ii) the fitted values of the model, (iii) the observed data, (iv and v) the results of the residuals normality and heterogeneity tests, and (vi) the log-transformation function used. If the argument compare = TRUE is used in lin_pow, a 7th element is returned that contains the parameter values from the non-linear power model.

For a 'sars' object of Type 'threshold', a list with three elements is returned: (i) the information criterion used to order the ranked model summary table (currently just BIC), (ii) a model summary table (models are ranked using BIC), and (iii) details of any axes log-transformations undertaken. Note that in the model summary table, if log-area is used as the predictor, the threshold values will be on the log scale used. Thus it may be preferable to back-transform them (e.g. using exp(th) if natural logarithms are used) so that they are on the scale of untransformed area. Th1 and Th2 in the table are the threshold value(s), and seg1, seg2, seg3 provide the number of datapoints within each segment (for the threshold models); one-threshold models have two segements, and two-threshold models have three segments.

Examples

data(galap)
#fit a multimodel SAR and get the model table
mf <- sar_average(data = galap, grid_start = "none")
summary(mf)
summary(mf)$Model_table
#Get a summary of the fit of the linear power model
fit <- lin_pow(galap, con = 1, compare = TRUE)
summary(fit)

sars documentation built on Dec. 28, 2022, 2:38 a.m.