binomial family argument

The big difference between running a linear regression with lm() and running a logistic regression with glm() is that you have to set glm()'s family argument to binomial. binomial() is a function that returns a list of other functions that tell glm() how to perform calculations in the regression. The two most interesting functions are linkinv and linkfun, which are used for transforming variables from the whole number line (minus infinity to infinity) to probabilities (zero to one) and back again.

A vector of values, x, and a vector of probabilities, p, are available.

Deze oefening maakt deel uit van de cursus

Intermediate Regression in R

Cursus bekijken

Oefeninstructies

Examine the structure of the binomial() function. Notice that it contains two elements that are functions, binomial()$linkinv, and binomial()$linkfun.
Call binomial()$linkinv() on x, assigning to linkinv_x.
Check that linkinv_x and plogis() of x give the same results with all.equal().
Call binomial()$linkfun() on p, assigning to linkfun_p.
Check that linkfun_p and qlogis() of p give the same results.

Praktische interactieve oefening

Probeer deze oefening eens door deze voorbeeldcode in te vullen.

# Look at the structure of binomial() function
___

# Call the link inverse on x
linkinv_x <- ___

# Check linkinv_x and plogis() of x give same results 
___

# Call the link fun on p
linkfun_p <- ___

# Check linkfun_p and qlogis() of p give same results  
___

Code bewerken en uitvoeren