What is the Pr(>|z|) telling me in a binomial summary

Hi there,
I'm wondering if anyone knows what Pr(>|z|) and the z value means in the summary of a binomial glm of a model represents. Here is the print out and code: Setp is the response variable of settlement of larvae, PLD is a factor of time and habitat is where the larvae settled.

BL1 <- glm(Setp ~ PLD * Habitat, family = binomial,
data = BL)
summary(BL1)
Call:
glm(formula = Setp ~ PLD * Habitat, family = binomial, data = BL)

Deviance Residuals:
Min 1Q Median 3Q Max
-0.50751 -0.34671 -0.07356 0.16767 0.65171

Coefficients:
Estimate Std. Error z value Pr(>|z|)
(Intercept) -0.4578 0.5132 -0.892 0.3723
PLD60 -1.6939 0.9660 -1.754 0.0795 .
PLD240 -1.7866 0.9923 -1.800 0.0718 .
HabitatDark Smooth -0.3895 0.7490 -0.520 0.6031
HabitatLight Rough -1.9401 1.0400 -1.866 0.0621 .
HabitatLight Smooth -2.4022 1.2178 -1.973 0.0486 *
PLD60:HabitatDark Smooth 0.5567 1.3487 0.413 0.6798
PLD240:HabitatDark Smooth -0.3106 1.6118 -0.193 0.8472
PLD60:HabitatLight Rough 1.7473 1.5919 1.098 0.2724
PLD240:HabitatLight Rough 1.4030 1.7148 0.818 0.4133
PLD60:HabitatLight Smooth 2.3095 1.6953 1.362 0.1731
PLD240:HabitatLight Smooth 2.0075 1.7913 1.121 0.2624

Signif. codes: 0 ‘’ 0.001 ‘’ 0.01 ‘’ 0.05 ‘.’ 0.1 ‘ ’ 1

(Dispersion parameter for binomial family taken to be 1)

Null deviance: 33.011  on 191  degrees of freedom

Residual deviance: 18.402 on 180 degrees of freedom
AIC: 78.458

Number of Fisher Scoring iterations: 5

I'm trying to determine if there are significant differences because the boxplots of the raw data showed there was. Does the Pr(>|z|) need to be <0.05 to be significant?

Thanks in advance

Here is a breakdown of the values shown for one of your parameters, HabitatLightSmooth.

                    Estimate  Std Error  z value  Pr(>|z|)
HabitatLight Smooth -2.4022   1.2178     -1.973   0.0486

Estimate/Std Error = -2.4022/1.2178 = -1.972573 = z value

Calculate the fraction of the normal distribution that is beyond the z value.
The multiplication by two accounts for the absolute value of z.
It is the area below -1.973 and above 1.973.

pnorm(-1.972573) * 2 = 0.04854423 = Pr(>|z|)  

The z value tells you how many standard errors lie between zero and the estimated value of the parameter. Pr(|z|) tells you how much of the normal distribution is farther out, on either tail, than the absolute value of your z score.
A p value (Pr(|z|)) of 0.05 should not be treated as a bright line of significance. If your data had been very slightly different, giving a estimate of -2.366 instead of -2.4022, the p value of HabitatLightSmooth would have been 0.052. That parameter would not then be "insignificant" if it is "significant" now.

This topic was automatically closed 21 days after the last reply. New replies are no longer allowed.

If you have a query related to it or one of the replies, start a new topic and refer back with a link.