GWR4.09 is now available and supersedes any previous versions. It can be downloaded from the link at the bottom of the page. This version of GWR is a new release of the Windows application software tool for modelling spatially varying relationships among variables by calibrating Geographically Weighted Regression (GWR) and Geographically Weighted Generalised Linear Models (GWGLM) with their semiparametric variants.
The GWR4 software was developed and programmed by Professor Tomoki Nakaya of the Department of Geography, Ritsumeikan University, Kyoto, Japan. The main new features of GWR4 are: Fitting GWR models: SGWRWIN (GWR4.0) enabling the user to fit a range of GWR models including conventional Gaussian models as well as its extensions based on the framework of generalised linear modelling including geographically weighted Poisson and logistic regression models.
Compared with the previous version (GWR3.x), a wider range of options related to GWR modelling like kernel function specification & bandwidth selection are available as the followings. 1. Different combinations of fixed-adaptive and Gaussian/bi-square functions are available. 2. Four criteria (AICc, AIC, BIC/MDL, CV) can be chosen as a model selection criterion. 3. In addition to the golden section search routine, an interval search routine, an exhaustive search by a regular step between pre-specified upper and lower limits, is included as an option of bandwidth optimisation method.
For conventional GWR modelling, SGWRWIN (GWR4.0) generally provides results that are consistent with that of the previous version in most of estimates and diagnostic indicators, though a few corrections are made about calculation methods of local diagnostic statistics including local sigma and local R square. Semiparametric GWR: New functions have been provided to fit semiparametric (partial linear) GWR models in which you may mix globally fixed and locally varying explanatory variables simultaneously. The function can be applied to a range of generalized linear modelling including Gaussian, Poisson and logistic regression. Using the scheme of semiparametric modelling, a new test of geographical variability on geographically varying coefficients is now enabled. It is also possible to use variable selection routines by which variables are automatically selected as either fixed or varying terms by recursive model comparisons.
New interface: The user interface has been largely rebuilt. A new tabbed interface is introduced so that a modelling session is intuitively proceeded by a step-by-step manner. Datasets and geographically listwise results can be viewed in separate spread-sheet like windows. A wider range of file types can be used for an input data file (space, comma, tab separated text formats & dbase IV format). Also, areal key field can be integrated into an output of GWR modelling so that you may join your output CSV file to a GIS attribute table by the key field for mapping the result in a GIS environment. A set of modelling specifications for the software can be stored in a file and used for a batch mode without window interfaces at a command line of MS-Windows.
Download a version of GWR4 4.09 for your system (Windows 64 bit, or Windows 32 bit/Windows XP 32 bit) together with example data from here. Unzip the file to a convenient folder and install the appropriate GWR_setup application for your system. You will also find the latest version of the manual in the zipfile.
Problems with the installation? If you are having problems installing the version from the link above, then there is another 64 bit installer which can be downloaded from here.