itsa.postest: Post-estimating ITSA models

Description Usage Arguments Details Value

View source: R/itsa.postest.R

Description

Function to produce post-estimation plots from the itsa.model function.

Usage

1
2
3
4
5
6
7
itsa.postest(
  model = NULL,
  no.plots = FALSE,
  alpha = 0.05,
  bootstrap = TRUE,
  print = TRUE
)

Arguments

model

define ITSA model results object.

no.plots

logical, specify whether function should present post-estimation plots (default is FALSE).

alpha

define alpha level for test statistics, default is 0.05.

bootstrap

logical, declare if itsa.model was bootstrapped for plot of these results.

print

logical, specify whether the main model result should be printed to the console.

Details

This function runs and reports post-estimation tests on fits from the itsa.model function, and generates four plots.

Main tests are whether two key ANCOVA assumptions are met, and an additional autocorrelation test for the time series framework.

The Shapiro-Wilks test examines the residuals from the fitted model for abnormality. A p-value less than alpha indicates abnormal residuals. In this instance, the user should consider replicating the its.analysis model design using a non-parametric ANCOVA equivalent.

The Levene's Test makes sure that there are equal variances between the treated groups. A p-value less than alpha indicates heterogeneous variances.

A QQ-Norm and Boxplot are generated with the test results overlaid (respectively), with a Residual v Fitted and Autocorrelation Function Plot also generated.

The results of bootstrap estimations in itsa.model will be plotted, unless argument is switched to FALSE. See plot.boot help file for assistance on interpreting these plots.

Default is to generate plots and summary table, but plots may be overridden using no.plots argument.

Default alpha value for post-estimation statistics is 0.05, test results will suggest potential presence of problems at higher values (and also at higher levels relative to a user-defined alpha), but user discretion is needed (examined in tandem with the Residuals v Fitted plot).

See 'itsa.model' documentation for further information.

Value

Returns a list containing post-estimation test results and four plots (plus one additional if bootstrapping applied) which are also sent directly to the plot tab.


patrick-eng/ITS.analysis documentation built on Jan. 5, 2021, 2:46 a.m.