EXERCISE 6 - ANSWER

Well, as it is reasonable to think that the size of the residence can affect its monthly cost, it is also reasonable to think that the number of people in a family will determine the size of the house that it chooses. Is this true in the Saldru dataset? If so, an additional person is likely to make a family acquire how many more rooms (on average)? Please show another graph.

The variables of interest here are rooms_to (again the number of rooms in a residence) and hhsizem (number of people in a household). Again, these variables are measured at the household level, so it is important to reduce the dataset down to one observation per household. The best way to do this is to drop everyone who isn't the respondent to the survey by using the command:

keep if pers_res==1

Now use the following regression command to generate this table:

regress rooms_to hhsizem

Source	SS	df	MS		Number of obs	990
					F(  1,   988)	59.39
Model	345.241841	1	345.241841		Prob > F	0
Residual	5743.46624	988	5.81322494		R-squared	0.0567
					Adj R-squared	0.0557
Total	6088.70808	989	6.1564288		Root MSE	2.4111

rooms_to	Coef.	Std. Err.	t	P>t	[95% Conf.	Interval]
hhsizem	0.2011084	0.0260962	7.706	0	0.1498981	0.2523188
_cons	3.196684	0.1431271	22.335	0	2.915815	3.477552

Because the t-stat for hhsizem (7.706) is greater than 1.96, it is safe to say that the number of people in a household significantly affects the number of rooms the family chooses for its dwelling. In fact, the coefficient of .2011 for hhsizem suggests that an additional person would make a family rent or purchase one-fifth of another room. What is a fifth of a room, you ask? Said differently, after an increase of five more people, an average family would acquire another room.

To graph this relationship, you must first "predict" the values of the regression. We called our predicted variable "roomshat" but any name would have sufficed.

predict roomshat

With the following graphing command, you should get this resulting graph:

graph rooms_to roomshat hhsizem, connect(.s) symbol (oi) ylabel xlabel