Question

I know how to produce the plots using leveragePlot(), but I can not find a way to produce a statistic for leverage for each observation like in megastat output.

Answer 1

I think you’re looking for the hat values.

Use hatvalues(fit). The rule of thumb is to examine any observations 2-3 times greater than the average hat value. I don’t know of a specific function or package off the top of my head that provides this info in a nice data frame but doing it yourself is fairly straight forward. Here’s an example:

fit <- lm(hp ~ cyl + mpg, data=mtcars) #a fake model

hatvalues(fit)

hv <- as.data.frame(hatvalues(fit))
mn <-mean(hatvalues(fit))
hv$warn <- ifelse(hv[, 'hatvalues(fit)']>3*mn, 'x3',
   ifelse(hv[, 'hatvalues(fit)']>2*mn, 'x3', '-' ))

hv

For larger data sets you could use subset and/or orderto look at just certain values ranges for the hat values:

subset(hv, warn=="x3")
subset(hv, warn%in%c("x2", "x3"))
hv[order(hv['hatvalues(fit)']), ]

I actually came across a nice plot function that does this in the book R in Action but as this is a copyrighted book I will not display Kabacoff’s intellectual property. But that plot would work even better for mid sized data sets.

Here is a decent hat plot though that you may also want to investigate:

plot(hatvalues(fit), type = "h")