One way to run such a regression in R Studio is as follows:

ff <- fin$FEES

af <- fin$ASSETS

LFF <- log(ff)

LFA <- log(af)

B4F <-factor(fin$BIG4)

resultsf <-lm(LFF~LFA+B4F) summary(resultsf)

Assignment:

What is the predicted audit fees for a bank with $500 million in assets?

How does this number compare with the predicted fee if we used the full dataset (i.e., the entire sample of 4426 observations without deleting the non-financial companies)? (Recall that we did this earlier, and you have the regression results from the full sample.)

Bonus question: How does the predicted audit fee for a bank with $500 million in assets compare with the predicted audit fees for a non-financial firm with $500 million in assets? What are the lessons to learn from such a comparison?