使用glmulti函数提取p值

Question

I am using the package glmulti to find the "best" model for

mo <- glm(a ~ x + y + z, data=data, family=binomial)
test <- glmulti(mo, family = binomial, level=1, crit="aicc")

However, I am more interested in the p-values of x in all models (to find the best p-value). How can I extract them from all models?

Answer 1

test outputs the best model, but gmulti allows you to see all the models. You can see all other models with their weights and aicc values with weightable

weightable(test)

Here is a loop through the resultset, build glm model with as.formula() and extract the p-values of x (if present) from the summaries.

library(glmulti)

iris2 <- iris
iris2$virginica <- iris$Species == "virginica"

mo = glm(virginica ~ Sepal.Length + Sepal.Width + Petal.Length + Petal.Width, data=iris2, family="binomial")

test <- glmulti(mo, family = binomial, level=1, crit="aicc")

tbl <- weightable(test)
nrows <- nrow(tbl)

for (i in 1:nrows) {
    formulaTest <- as.character(tbl[i,]$model)
    model = glm(as.formula(formulaTest), data=iris2, family="binomial")

    print(paste("Formula: ", formulaTest))
    print(summary(model)$coefficients[,4])
}

[1] "Formula:  virginica ~ 1 + Sepal.Width + Petal.Length + Petal.Width"
 (Intercept)  Sepal.Width Petal.Length  Petal.Width 
  0.03522621   0.07853563   0.04033382   0.04534974 
[1] "Formula:  virginica ~ 1 + Sepal.Length + Sepal.Width + Petal.Length + Petal.Width"
 (Intercept) Sepal.Length  Sepal.Width Petal.Length  Petal.Width 
  0.09720282   0.30318813   0.13585218   0.04653593   0.06052811 
[1] "Formula:  virginica ~ 1 + Sepal.Length + Petal.Length + Petal.Width"
 (Intercept) Sepal.Length Petal.Length  Petal.Width 
  0.02770914   0.07149206   0.02891997   0.02830069 
...

If you want to employ advanced model selection, you could even combine top models, as described here: http://www.noamross.net/blog/2013/2/20/model-selection-drug.html