spdep (version 0.5-74)

lm.LMtests: Lagrange Multiplier diagnostics for spatial dependence in linear models

Description

The function reports the estimates of tests chosen among five statistics for testing for spatial dependence in linear models. The statistics are the simple LM test for error dependence (LMerr), the simple LM test for a missing spatially lagged dependent variable (LMlag), variants of these robust to the presence of the other (RLMerr, RLMlag - RLMerr tests for error dependence in the possible presence of a missing lagged dependent variable, RLMlag the other way round), and a portmanteau test (SARMA, in fact LMerr + RLMlag). Note: from spdep 0.3-32, the value of the weights matrix trace term is returned correctly for both underlying symmetric and asymmetric neighbour lists, before 0.3-32, the value was wrong for listw objects based on asymmetric neighbour lists, such as k-nearest neighbours (thanks to Luc Anselin for finding the bug).

Usage

lm.LMtests(model, listw, zero.policy=NULL, test="LMerr", spChk=NULL, naSubset=TRUE)
## S3 method for class 'LMtestlist':
print(x, ...)
## S3 method for class 'LMtestlist':
summary(object, p.adjust.method="none", ...)
## S3 method for class 'LMtestlist.summary':
print(x, digits=max(3, getOption("digits") - 2), ...)

Arguments

model

an object of class lm returned by lm, or optionally a vector of externally calculated residuals (run though na.omit if any NAs present) for use when only "LMerr" is chosen; weights and offsets should not be used in t

listw

a listw object created for example by nb2listw, expected to be row-standardised (W-style)

zero.policy

default NULL, use global option value; if TRUE assign zero to the lagged value of zones without neighbours, if FALSE assign NA

test

a character vector of tests requested chosen from LMerr, LMlag, RLMerr, RLMlag, SARMA; test="all" computes all the tests.

spChk

should the data vector names be checked against the spatial objects for identity integrity, TRUE, or FALSE, default NULL to use get.spChkOption()

naSubset

default TRUE to subset listw object for omitted observations in model object (this is a change from earlier behaviour, when the model$na.action component was ignored, and the listw object had to be subsetted by hand)

x, object

object to be printed

p.adjust.method

a character string specifying the probability value adjustment (see p.adjust) for multiple tests, default "none"

digits

minimum number of significant digits to be used for most numbers

...

printing arguments to be passed through

Value

• A list of class LMtestlist of htest objects, each with:
• statistic
  the value of the Lagrange Multiplier test.
• parameter
  number of degrees of freedom
• p.value
  the p-value of the test.
• method
  a character string giving the method used.
• data.name
  a character string giving the name(s) of the data.

Details

The two types of dependence are for spatial lag $\rho$ and spatial error $\lambda$:

$$\mathbf{y} = \mathbf{X \beta} + \rho \mathbf{W_{(1)} y} + \mathbf{u},$$ $$\mathbf{u} = \lambda \mathbf{W_{(2)} u} + \mathbf{e}$$

where $\mathbf{e}$ is a well-behaved, uncorrelated error term. Tests for a missing spatially lagged dependent variable test that $\rho = 0$, tests for spatial autocorrelation of the error $\mathbf{u}$ test whether $\lambda = 0$. $\mathbf{W}$ is a spatial weights matrix; for the tests used here they are identical.

References

Anselin, L. 1988 Spatial econometrics: methods and models. (Dordrecht: Kluwer); Anselin, L., Bera, A. K., Florax, R. and Yoon, M. J. 1996 Simple diagnostic tests for spatial dependence. Regional Science and Urban Economics, 26, 77--104.

See Also

lm

Examples

data(oldcol)
oldcrime.lm <- lm(CRIME ~ HOVAL + INC, data = COL.OLD)
summary(oldcrime.lm)
res <- lm.LMtests(oldcrime.lm, nb2listw(COL.nb), test=c("LMerr", "LMlag",
  "RLMerr", "RLMlag", "SARMA"))
summary(res)
lm.LMtests(oldcrime.lm, nb2listw(COL.nb))
lm.LMtests(residuals(oldcrime.lm), nb2listw(COL.nb))