------------------------------------------------------------------------------- help for splagvar -------------------------------------------------------------------------------


splagvar --- Generates spatially lagged variables, constructs the Moran scatter plot, and calculates global Moran's I statistics.

+--------------------+ ----+ Table of Contents +-----------------------------------------------

Syntax General description Description of the options Examples References Author information Citation



splagvar [varlist1] [if] [in], [wname(weights_name) wfrom(Stata|Mata) ind(varlist2) Other_options]

options Description ------------------------------------------------------------------------- Options wname(weights_name) indicate the name of the spatial weights matrix to be used

wfrom(Stata|Mata) indicate source of the spatial weights matrix ind(varlist2) request spatially lagged explanatory variables

order(#) indicate the lag order for the spatially lagged explanatory variables to be created

equation(#) generate spatially lagged variables for a specific equation

plot(varname) construct the Moran scatter plot

title(title_info) indicate a title for the Moran scatter plot

saving(filename, ...) save the graph to filename

note(note_info) provide notes about the graph

reps(#) set the number of random permutations

seed(#) specify a seed number

moran(varname) request Moran's I statistics

qvar(varlist3) generate quasi-instrumental variables

qname(newvarlist) provide names for the quasi-instrumental variables

replace overwrite existing spatially lagged variables

favor(speed|space) favor speed or space when calculating the spatially lagged variables and the Moran's I statistics


+-------------+ ----+ Description +------------------------------------------------------

splagvar generates spatially lagged variables, constructs the Moran scatter plot, and calculates global Moran's I statistics to test for the presence of spatial dependence. The Moran's I p-value displayed on the Moran scatter plot is calculated using a random permutation procedure. Optionally, quasi-instruments can be generated. Unless quasi-instruments are requested, splagvar requires a spatial weights matrix which must have been created using spwmatrix or any other commands that generate spatial weights. The purpose of splagvar is to facilitate estimation by spatial two-stage least squares and generalized method of moments of spatial econometrics models in Stata (see the example section).

While generated spatially lagged dependent variables are prefixed with wy_, generated first, second, and third order spatially lagged explanatory variables are prefixed with wx_, w2x_, and w3x_ respectively. The names of the supplied variables make up the remaining portions. If option equation(#) is specified, then the spatially lagged variables are prefixed with wy#_, wx#_, w2x#_, and w3x#_, where # stands for the equation number.

splagvar requires Stata 10.1 or higher.

+---------+ ----+ Options +----------------------------------------------------------

wname(weights_name) specifies the name of the spatial weights matrix to be used.

wfrom(Stata | Mata) indicates whether the spatial weights matrix is a Stata matrix loaded in memory or a Mata file located in the working directory. When generated by spwmatrix the spatial weights matrix will exist as a Stata matrix or as a Mata file.

wname() and wfrom() are required, unless qvar() and qname() are specified.

ind(varlist2) specifies a list of explanatory variables whose spatial lags need to be taken. Variables deemed dependent or endogenous must be supplied in varlist1.

order(#) specifies the spatial lag order (up to 3) for the spatially lagged explanatory variables to be created. By default, first order spatial lags are calculated. For dependent or endogenous variables, only first order spatial lags may be generated.

equation(#) indicates the number of the equation for which spatially lagged variables need to be calculated when generating variables for multiple equations. For example, specifying equation(1) instructs Stata to generate spatially lagged variables for the first equation. This option is crucial for estimating a spatial simultaneous equation model where a different spatial weights matrix is used for each equation in Stata.

plot(varname) constructs the Moran scatter plot for the dependent variable varname listed in varlist1. On the plot are reported Moran's I statistics and the associated p-value based on a random permutation procedure.

title(title_info) specifies a title for the Moran scatter plot.

note(note_info) provides notes associated with the graph.

saving(filename, ...) saves the graph to filename. Specifying the suboption replace with the option saving() will replace the graph if it already exists.

seed(#) sets the random-number seed, which is defaulted to seed(042009). This option is useful to ensure replicability of the results.

reps(#) specifies the number of random permutations to be performed. The default is reps(999).

moran(varname) requests that global Moran's I statistics calculated under the assumptions of normal approximation and randomization be displayed. Again, varname must be from varlist1.

N.B.: If varlist1 is not specified, none of the options plot(), saving(), title(), moran(), note(), reps(), and seed() may be specified.

qvar(varlist3) specifies the variables to be used in generating quasi-instrumental variables.

qname(newvarlist) specifies a list of variable names for the generated quasi-instrumental variables coded 1, 0, and -1 depending on whether or not the values of the corresponding variables specified with qvar() are in the upper, middle or lower third of values when placed in rank order (see Fingle and Le Gallo, 2008).

replace overwrites existing spatially lagged or quasi-instrument variables.

favor(speed|space) instructs splagvar to favor speed or space when calculating the spatially lagged variables or the Moran's I statistics. favor(speed) is the default. This option provides a tradeoff between speed and memory use. See [M-3] mata set.

+----------+ ----+ Examples +---------------------------------------------------------

1) Create a spatial lag for the dependent variable, construct the Moran scatter plot, and request Moran's I statistics

. splagvar povrate, wname(C:\data/spweight) wfrom(Mata) plot(povrate) moran(povrate)

2) Create a spatially lagged dependent variable and request Moran's I statistics and p-value under the assumption of normal approximation and randomization

. splagvar crimerate, wname(spweight) wfrom(Stata) moran(crimerate)

3) Create a spatially lagged dependent variable and first and second order spatially lagged independent variables

. splagvar houseval, wname(spweight1) wfrom(Mata) ind(income population education) order(2)

Now estimating a spatial lag model by spatial 2SLS and GMM using WX and (W^2)X as instruments has never been easier.

. ivregress 2sls houseval (wy_houseval=wx_* w2x_*) income population education, vce(robust)

. ivregress gmm houseval (wy_houseval=wx_* w2x_*) income population education

Other endogenous variables can be added to the model provided that instruments are available.

. ivregress gmm houseval (wy_houseval endogvar=wx_* w2x_* endogvar_instr) income population education

Education can be considered endogenous if you can find a good instrument.

. ivregress gmm houseval (wy_houseval education=wx_* w2x_* education_instr) income population

Iterative GMM can be more efficient than two-step GMM.

. ivregress gmm houseval (wy_houseval education=wx_* w2x_* education_instr) income population, igmm

Note: If you want to use the lags of an instrumental variable as instruments, then you need to include it in varlist2

4) Create a quasi-instrumental variable to be used as an instrument for education

. splagvar, qvar(education) qname(educ_qinst)



Anselin, L. 2005. Exploring Spatial Data with GeoDaTM : A Workbook. Available a > t: http://www.csiss.org/clearinghouse/GeoDa/geodaworkbook.pdf

Anselin, L. 2007. "Spatial Econometrics". In T. C. Mills and K. Patterson (Eds) > . Palgrave Handbook of Econometrics. Vol 1, Econometric Theory. New York: Palgrave MacMillan, pp. 901-969.

de Smith, M.J., M.F. Goodchild, and P.A. Longley. 2007. Geospatial Analysis: A > comprehensive Guide to Principles, Techniques, and Software Tools. Matador: Leicester, UK http://www.spatialanalysisonline.com

Fingleton. B. and J. Le Gallo. 2008. "Estimating Spatial Models with Endogenous > Variables, a Spatial Lag and Spatially Dependent Disturbances: Finite Sample Properties ", Papers in Regional Science > 87(3): 319-339.

Gould, W. 2006. "Mata Matters: Creating New Variables—Sounds Boring, Isn't". Th > e Stata Journal 6: 112-123. Available from http://www.stata-journal.com/article.html?article=pr0021

Klotz, S. 2004. Cross-sectional Dependence in Spatial Econometrics Models: with > an Application to German Start-up Activity Data. Transaction Publishers, Piscataway, NJ.


P. Wilner Jeanty, The Kinder Institute for Urban Research/Hobby Center for the Study of Texas, Rice University, Houston, Texas

Email to pwjeanty@rice.edu

N.B.: Previous versions of splagvar were written when the author was a Research Economist with the Dept. of Agricultural, Environmental, and Development Economics, The Ohio State University


Thanks for citing splagvar as follows:

Jeanty, P.W., 2010. splagvar: Stata module to generate spatially lagged variabl > es, construct the Moran scatter plot, and calculate global Moran's I statisti > cs Available from http://ideas.repec.org/c/boc/bocode/s457112.html.

Also see

Online: spwmatrix, spatgsa, spatcorr (if installed)