gq.test: Goldfeld-Quandt Test

Description

Goldfeld-Quandt test for heteroskedastic errors. The object of test results returned by this command can be plotted using the plot() function.

Usage

gq.test(
  mod,
  data = list(),
  split = 0.5,
  omit.obs = 0,
  ah = c("increasing", "unequal", "decreasing"),
  order.by = NULL,
  sig.level = 0.05,
  details = FALSE,
  hyp = TRUE
)

Value

A list object including:

`hyp`	character matrix of hypotheses (if `hyp = TRUE`).
`results`	a data frame of basic test results.
`hreg1`	matrix of regression results in Group I.
`stats1`	additional statistic of regression in Group I.
`hreg2`	matrix of regression results in Group II.
`stats2`	additional statistic of regression in Group II.
`nulldist`	type of the Null distribution with its parameters.

Arguments

mod: estimated linear model object or formula. If only a model formula is passed then the data argument must be specified.
data: if mod is a formula then the corresponding data frame has to be specified.
split: partitions the data set into two groups. If <= 1 then split is a percentage value such that T*split observations are in the first partition. If split >= 1 it is interpreted as the index of the partitioning observation, i.e. the number of observations in the first group.
omit.obs: the number of central observations to be omitted. Might increase the power of the test. If <= 1 then split is the percentage value of all observations, otherwise it is interpreted as absolute number.
ah: character string specifying the type of the alternative hypothesis: "increasing" (variance increases from group 1 to group 2), "decreasing" (variance decreases from group 1 to group 2), "unequal" (variances are unequal between the groups). The default is to test for increasing variances.
order.by: either a vector z or a formula with a single explanatory variable like ~ z. The observations in the model are ordered by the size of z. If set to NULL (the default) the observations are assumed to be ordered.
sig.level: significance level. Default value: sig.level = 0.05.
details: logical value indicating whether specific details about the test should be returned.
hyp: logical value indicating whether the Hypotheses should be returned.

References

Goldfeld, S.M. & Quandt, R.E. (1965): Some Tests for Homoskedasticity. Journal of the American Statistical Association 60, 539-547.

Examples

Run this code

## 5 observations in group 1 with the hypothesis that the variance of group 2 is larger
gq.test(rent ~ dist, split = 5, ah = "increasing", data = data.rent)

## Ordered by population size
eu.mod <- ols(expend ~ pop + gdp + farm + votes + mship, data = data.eu)
results <- gq.test(eu.mod, split = 13, order.by = data.eu$pop, details = TRUE)
results

plot(results)