Exam 3 Cumulative Review Flashcards
after a simple p and q regression, how do we find the elasticity of something at a certain point?
(p/q) x (coefficient)
does elasticity along a linear demand curve change?
yes, depending on where you are on the curve
whats the most basic way to evaluate curvilinear data?
log-log regression!
convert all the data to Ln’s
how do convert a basic logged q and p regression back?
- take the log of the intercept
- take the coefficient (which is the elasticity) and use that as the exponent for p
- multiply both of these
q = ln (coefficient) x p ^ (elasticity) Q_demanded = 23,156* P(-.873)
what happens to the interpretation of a log-log?
taking the log of both variables changes the interpretation to percentages
a certain percent change in X is associated with a certain percentage change in Y
what is the value of beta if our curve is convex and starts near the origin?
beta will be greater than 1
what is the value of our beta if our curve is concave, originating near the origin?
beta is between 0 and 1
what is the value of beta if our curve starts high and is convex down?
beta is less than 0
log linear regression
- convert DV variable to logs
- leave the IV as they are
- interpretation? changes to the dependent variable are interpreted as percentages, not units
interpretation of:
Ln(selling price) = 6.21 - 0.08*(age)
it’s a log linear, so …
a one year increase in the age of the house translates to an 8% increase in the selling price of the house
can you compare the R^2 or the adjusted R^2’s between a linear and log-linear regression?
NO, because the DV’s are two different things
linear log regression
convert the IV to logs
-leave the DV as it is
changes in the IV are not interpreted as percentages, not units
interpretation of:
number of dining out experiences = -37.9 + 11.33*(LnIncome)
this is a linear log soooo….
a 9% increase in income translates to (9*.11 or .99) or 1 time increase in dining out per month
indifference curves
curves that indicate pairs of things you like equally well
-come from utility functions…
if given this, how would you create an indifference curve?:
U = 3.2(# good X)^(.46)(# Good Y)^(.64)
you find the different combinations that will give you an x amount of utility, and graph all of those points to make a nice little curve
how do you get a utility function?
data from market research
- ask consumers to rate any combination of goods
- ex: beers, pizzas, satisfaction score
estimated utility function formula
Ln(utility) = ln(a) + (b)(Ln_beer) + (c)(Ln_pizza)
now convert it back
utility = (exp(.635))beer^(.502)pizza^(.163)
OR
utility = 1.89beer^(.502)pizza^(.163)
Hedonic regression
- dissecting a product into it’s characteristics
- used to understand how consumers view the trade-offs among those characteristics
- fantastic use of utility functions
- often used in real estate
example of Hedonic regression
P = .202429*Bed^(.155)Bath^(.2843)Squarefeet^(.334)
omitted variable bias
a constant tacked onto the end of a regression equation to account for unaccounted for variables
panel data
- cross section and time series data combined!
- data varies across entities and across time
- diff markets and years
- diff states and diff months
what is one way we account for time in different cities in a panel data regression?
- instead of compared the price and quantity demanded of all of the cities in each year, we look at the average price and average quantity demanded for EACH city
- we then take “P - avg. P” And ‘Q - Avg. Q” for each city and for each year
- regress data found in step 2
whats another easier way that we can run a regression with panel data?
- create a dummy variable for all cities except one (the intercept will be the city without a dummy)
- create another binary dummy variable for time
fixed effects regression
- when we hold constant the average effects of each city
- create state-specific binary variables to capture fixed effects
time fixed effects regression
just running a dummy variable for time now
how can we test, with a time fixed effects regression, if the coefficients are indistinguishable from zero?
you know dis!
(RSSrestricted-RSSunrestricted/q)
__________________________________
(RSSunrestricted / n-k-1)
simply use your time fixed effects regression as your unrestricted and your very first regression with no effects as your restricted
whats the command to run timed fixed effects and fixed effects regressions in STATA
xi: regress y x1 x2 x3 x4 i.x5 i.x6
