[R] Correct coefficients from treatment contrasts?

[David Winsemius]

On Sep 6, 2010, at 4:03 AM, B W wrote:

<Snipped out formatting detritus and added back many missing speces.>

> ->Hello,I am trying to take the information from the summary of my  
> best fit logisticregression model for the occurrence of a high  
> elevation plant spp. and create the appropriate equation that will  
> calculate probability of occurrence, given the data. My predictors  
> include both continuous variables (slope and a second  
> orderpolynomial of elevation) and a discrete variable for aspect  
> (warm and cool). I have left unchanged the default contrasts option,  
> so I believe that thefollowing coefficients were created using  
> treatment contrasts.  My question how can I take this summary output  
> and create the logistic equation that will allow me to calculate  
> probability of occurrence. My interests are touse this to spatially  
> display this info in a GIS environment.

I think you should:

-- Read the Posting Guide where you should learn that this is a plain  
text mailing list and that you need to change the configuration of  
your mail client.

-- Read the help page and read other documentation regarding the use  
of the predict function.

> I have made adraft equation (shown below) that uses the coefficients  
> from this summaryoutput, but this appears to be incorrect – values  
> always return zeroprobabilities. Presumably I need to adjust the  
> values in some way – but I am unclear as to how to proceed.  
> Anyguidance would be appreciated!

>  >summary (

> Call:glm(formula= Po ~ Slope + poly(Elevation, 2) + Aspect_2, family  
> = quasibinomial) DevianceResiduals:     Min      1Q   Median        
> 3Q     Max  -1.0532  -0.4167 -0.2760  -0.1823   3.3376

> Coefficients:                      Estimate Std. Error t valuePr(>| 
> t|)    (Intercept)          -4.577707   0.222406 -20.583  < 2e-16 ***

> Slope                 0.039959   0.003593 11.121  < 2e-16 ***

> poly(Elevation,2)1   8.050898   5.601956  1.437   0.1508

> poly(Elevation,2)2 -37.694521   6.297806  -5.985 2.39e-09 ***

> Aspect_2w             0.429229   0.174760  2.456   0.0141 *  ---

You may get predictions at the original data points with:

pred < predict(model.Slope.Elevation.Aspect)

>  (1/ (1 +  exp(-1 * (-4.577707 + 0.039959*Slope + 8.050898 *  
> poly(Elevation, 2)1 + -37.694521 * poly(Elevation, 2)2 + 0.429229*  
> Aspect_2w)))))
>
> Brendan Wilson
> 2530 Alexis Road
> Shoreacres BC
> Canada  V1N 4P6
> Ph: 1.250.359.5905
>
>
>
> 	[[alternative HTML version deleted]]

David Winsemius, MD
West Hartford, CT