topic GWR Null Values in Spatial Statistics Questions https://community.esri.com/t5/spatial-statistics-questions/gwr-null-values/m-p/378659#M1231 <HTML><HEAD></HEAD><BODY><SPAN>I'm currently trying to run some GWR analyses in ArcGIS 10.0. I have four environmental variables for my dependant variable (species richness).</SPAN><BR /><SPAN>The relationships I'm running are not necessarily linear, so I have created second order polynomials to explain the variance for two of the environmental variables. This clears up a lot of the explained variance in OLS and GWR when all the variables are combined. However, I want to look at how the R2 and coefficients change for each of the individual variables at a very local scale. Three of the GWR and environmental variables work fine, but for temperature (second order polynomial) at a fine resolution, many of the cells contain null (or infinity) values. </SPAN><BR /><BR /><SPAN>I have read several discussion topics on this, which highlight the issue of multicollinearity in the environmental dataset. I get this, and the data does appear to perhaps be locally clustered. From reading these discussions, the two methods suggested to overcome this by creating new variables 1) mean - variable value 2) create a new range variable, have not solved my issue.</SPAN><BR /><BR /><SPAN>I was wondering if anyone had any advice on how to overcome this issue?</SPAN><BR /><BR /><SPAN>Cheers</SPAN><BR /><SPAN>Paul</SPAN></BODY></HTML> Thu, 03 May 2012 18:41:23 GMT PaulHolloway 2012-05-03T18:41:23Z GWR Null Values https://community.esri.com/t5/spatial-statistics-questions/gwr-null-values/m-p/378659#M1231 <HTML><HEAD></HEAD><BODY><SPAN>I'm currently trying to run some GWR analyses in ArcGIS 10.0. I have four environmental variables for my dependant variable (species richness).</SPAN><BR /><SPAN>The relationships I'm running are not necessarily linear, so I have created second order polynomials to explain the variance for two of the environmental variables. This clears up a lot of the explained variance in OLS and GWR when all the variables are combined. However, I want to look at how the R2 and coefficients change for each of the individual variables at a very local scale. Three of the GWR and environmental variables work fine, but for temperature (second order polynomial) at a fine resolution, many of the cells contain null (or infinity) values. </SPAN><BR /><BR /><SPAN>I have read several discussion topics on this, which highlight the issue of multicollinearity in the environmental dataset. I get this, and the data does appear to perhaps be locally clustered. From reading these discussions, the two methods suggested to overcome this by creating new variables 1) mean - variable value 2) create a new range variable, have not solved my issue.</SPAN><BR /><BR /><SPAN>I was wondering if anyone had any advice on how to overcome this issue?</SPAN><BR /><BR /><SPAN>Cheers</SPAN><BR /><SPAN>Paul</SPAN></BODY></HTML> Thu, 03 May 2012 18:41:23 GMT https://community.esri.com/t5/spatial-statistics-questions/gwr-null-values/m-p/378659#M1231 PaulHolloway 2012-05-03T18:41:23Z