Generalised Linear Model in R

#1
Hello Talk Stats Members,

I am having a little trouble with interpreting the output from my model can anyone give a helping hand.

Ok so it is a Generalised Linear Model in RStudio. My outcome variable is either 0 or 1.

Here is the output summary from my final model:

Call:
glm(formula = outcome ~ Location + Size + Proximity + Location:Size +
Location:Size:proximity, family = binomial(link = probit),
data = shark2, control = glm.control(50))

Deviance Residuals:
Min 1Q Median 3Q Max
-1.5045 -0.6426 -0.4042 0.7594 1.8072

Coefficients:
Estimate Std. Error z value Pr(>|z|)
(Intercept) 21.166 12.298 1.721 0.0852 .
Location2 -12.222 8.318 -1.469 0.1417
Size -34.364 18.093 -1.899 0.0575 .
Proximity -37.200 16.259 -2.288 0.0221 *
Location2:Size 21.437 12.333 1.738 0.0822 .
Location1:Size:proximity 56.707 23.854 2.377 0.0174 *
Location2:Size:proximity 55.816 24.854 2.246 0.0247 *
---
Signif. codes: 0 ‘***’ 0.001 ‘**’ 0.01 ‘*’ 0.05 ‘.’ 0.1 ‘ ’ 1

(Dispersion parameter for binomial family taken to be 1)

Null deviance: 44.252 on 32 degrees of freedom
Residual deviance: 25.948 on 26 degrees of freedom
AIC: 39.948

Number of Fisher Scoring iterations: 1

and here is a Model summary:

Call: glm(formula = outcome ~ Location + Size + Proximity + Location:Size +
Location:Size:proximity, family = binomial(link = probit),
data = shark2, control = glm.control(50))

Coefficients:
(Intercept) Location2
21.17 -12.22
Size Proximity
-34.36 -37.20
Location2:Size Location1:Size:proximity
21.44 56.71
Location2:Size:proximity
55.82

Degrees of Freedom: 32 Total (i.e. Null); 26 Residual
Null Deviance: 44.25
Residual Deviance: 25.95 AIC: 39.95

Can anyone help me get my head around what I need to understand to make my write up? I can see that some effects and interactions are significant and they make sense but I am not sure what to do next with the output.

thank you for any guidance.

Darren.
 
#2
The key to interpret its results is the coefficients and P values of the independent variables:

Code:
Coefficients:
 Estimate Std. Error z value Pr(>|z|) 
 (Intercept) 21.166 12.298 1.721 0.0852 .
 Location2 -12.222 8.318 -1.469 0.1417 
 Size -34.364 18.093 -1.899 0.0575 .
 Proximity -37.200 16.259 -2.288 [B]0.0221 [/B]*
 Location2:Size 21.437 12.333 1.738 0.0822 .
 Location1:Size:Proximity 56.707 23.854 2.377 [B]0.0174[/B] *
 Location2:Size:Proximity 55.816 24.854 2.246 [B]0.0247[/B] *
The significant predictors are marked by an asterisk (I made them bold font). P value is shown below "Pr(>|z|)". Also R marks those smaller than 0.1 by a dot. The sign of coefficient (look for numbers below "Estimate") shows if the predictor is positively correlated or negatively correlated with Y (outcome).