Tobler is a web app that offers a intuitive step-by-step process for estimation of spatial econometric models, including a basic spatial visualization, creating of spatial weighting matrix and spatial correlation tests. Currently, it is possible to estimate OLS, SAR, SEM, SAC, SLX, SDM and SDEM models, with maximum likelihood and IV/GMM estimators. Estimates of impacts are provided, when applicable.

This app was entirely created with R using the Shiny library. Spatial data is handled by the RGDAL, cleangeo and leaflet libraries. Spatial dependence tests and spatial models are provided by the spdep and spatialreg library.

The name is a tribute to the geographer Waldo R. Tobler and his First Law of Geography: "Everything is related to everything else, but near things are more related than distant things."

This is a prototype. All results must be checked and validated by the users.

Developers: Raphael Saldanha (Fiocruz); Eduardo Almeida (UFJF).



Data upload

Load here your spatial data file including variables. Currently, we suggest the GML format.









Spatial Contiguity Matrix

Inverse Distance Matrix

K-Nearest Neighbors Matrix

Baumont (2004) procedure

This procedure will regress a model by OLS and test the residuals for spatial autocorrelation (Moran's Test) using several K-Nearest neighbors matrixes, varing k from 1 to the selected maximum k. The matrix with the higher Moran's I will be selected.

Baumont (2004) link.

This adapted procedure will create several spatial weights matrixes (Queen, Rook, Inverse Distance and K-Nearest Neighbors with k = 1, 5, 10, 15 and 20) and estimate several spatial models (SAR, SEM, SAC, SLX, SDM, SDEM) with those matrixes. The final matrix will be selected observing the maximum AIC obtained among all models.

Stakhovych-Bijmolt (2009) link.



Spatial autocorrelation



Moran's I scatter plot



Spatial models



OLS model

Ordinary Least Squares estimator for linear regression.

Estimation


                              

Moran's I for error term


                              

Residual map

Lagrange multiplier


                            

SAR (ML)

Spatial Autoregressive (SAR) model with maximum likelihood (ML) estimator.

Estimation


                              

Impacts


                              

Residual map

SAR (STSLS)

Spatial Autoregressive (SAR) model with Generalized Spatial Two Stage Least Squares (STSLS) estimator.

Estimation


                              

Impacts


                              

Residual map

SEM (ML)

Spatial Error Model (SEM) model with maximum likelihood (ML) estimator.

Estimation


                              

Residual map

SEM (GMM)

Spatial Error Model (SEM) model with Generalized Method of Moments (GMM) estimator.

Estimation


                              

Residual map

SAC (ML)

Spatial Autocorrelation Model (SAC) with maximum likelihood (ML) estimator.

Estimation


                              

Impacts


                              

Residual map

SAC (GSTSLS)

Spatial Autocorrelation Model (SAC) model with Generalized Spatial Two Stage Least Squares (GSTSLS) estimator.

Estimation


                              

Impacts


                              

Residual map

SLX (ML)

Spatial Lag X (SLX) model with maximum likelihood (ML) estimator.

Estimation


                              

Impacts


                              

Residual map

Spatial Durbin Model (SDM) with maximum likelihood (ML) estimator

SDM model with ML estimator.

Estimation


                              

Impacts


                              

Residual map

Spatial Durbin Error Model (SDM) with maximum likelihood (ML) estimator

SDEM model with ML estimator.

Estimation


                              

Residual map