How to display the exact p-values instead of < 0.001***?

Question

I conducted a logistic regression and multiple comparison

library(lme4)
model < glmer(y ~ Genger + Age + subject +
    (1 | ParticipantID), data = data, family = binomial(link = "logit"))

library(multcomp)
multicomp<-glht(model,mcp(Behavior="Tukey"))
summary(multicomp)

The output:

                              Estimate  Std. Error  z value Pr(>|z|)    
subject2 - subject1 ==  0 -2.19503726  0.66179761 -2.89596  < 0.001 ***

I want to display the exact p-values instead of < 0.001*** . What should I do?

I tried

options(digits=10)

but the output did not change.

Answer 1

You can change it like this:

library(lme4)
data(Orthodont,package="nlme")
model <-lmer(distance ~ Sex + age + (age|Subject), data=Orthodont)

library(multcomp)
multicomp<-glht(model, mcp(Sex = "Tukey"))
res <- summary(multicomp)

print(res)
#                   Estimate Std. Error z value Pr(>|z|) 
#Female - Male == 0  -2.1454     0.7575  -2.832  0.00462 **


coeftable <- do.call(cbind, res$test[c("coefficients", "sigma", "tstat", "pvalues")])
colnames(coeftable) <- c("Estimate", "Std. Error", "z value", "Pr(>|z|)")

printCoefmat(coeftable) 

#more digits (but still some protection against showing more digits than meaningful)
printCoefmat(coeftable, digits = 10) 

#no protection
print(coeftable)

#even more digits
options(digits = 20)
print(coeftable)

However, there is a reason why the exact p-value isn't printed. It is easily impacted by precision of floating-point numbers and also based on statistical assumptions which introduce uncertainty because they don't hold exactly. The multcomp developers are trying to protect you against interpreting meaningless numbers.