Exercise

Reading a Multivariate (Multiple) Regression Table

In our previous eGulf example, we assumed that age is the only determinant of a WePhone's sales price. However, sellers on eGulf also have "feedback scores," ranging from 0-100, that indicate the quality of their previous transactions with customers. It seems likely that a seller's feedback score could positively impact a WePhone's sales price; that is, customers might be willing to pay more for WePhones sold by reputable sellers. However, if reputable sellers tended to sell newer WePhones, this would confound the relationship between sales price and age.

We could consider overcoming this by throwing out respondents with high feedback scores that sold new WePhones, so that WePhone age and WePhone feedback score were no longer correlated. That might let us look at the effect of the feedback score in a natural experiment analysis, but throwing out those high scores might imbalance our sample and make that analysis invalid. However, with OLS regression, we can simply add seller feedback scores as another determinant of our independent variable, thereby controlling for its confounding effect on our outcome variable. This is what is shown in the table below.

Variable Coefficient (SE)
Age -55.61 (4.94)
Feedback 14.15 (2.55)
Intercept -707.23 (234.38)

As you can see, when controlling for seller score, age has an even larger negative effect on sales. Using the information in this table and the R console as a workspace, calculate the difference (in absolute terms) in the expected sales price for phone that is 2 year old with a feedback score of 80, versus a phone that is 4 years old with a feedback score of 90. You can use the R console as a scratch space to do your calculations.

Instructions

50 XP

Possible answers