Statistical tests Flashcards
what does the overidentification test for IV do?
tests for instrument validity
what is the test for IV instrument validity?
overidentification test
what is the null of the overidentification test?
all instruments are valid
what does it mean if you reject the null for the over id test?
one or more instruments are not valid or model was misspecified
true or false: if you reject the null for the iv test (that all instruments are valid), then you have proof that the instruments are valid
a lack of rejection should not be interpreted as confirmation of validity.
How do you test for whether you need to use IV (meaning you have an endogegnous regressor)?
Hausman test
What is the null of the Hausman test for IV?
you don’t need to use IV–variables are exogenous
wha does it mean if you reject null of Hausman test for IV?
explanatory variable is endogenous and IV is appropriate
what does it mean if you fail to reject null of Hausman test for IV?
You should use OLS, because IV is less efficient than OLS
what is hausman test in context of FE?
testing whether you should use FE or RE
what is the null of the hausman test for FE?
Null: RE assumptions hold, both estimators consistent but RE is efficient.
what should you do if you reject null of hausman test for FE?
RE assumptions do not hold and the RE estimator
is inconsistent–you should use FE
true or false: If you fail to reject null of hausman test for FE, you can use RE
true
What is the null for the f test in general?
coefficients are jointly zero
how do you test for weak IV instruments?
F test
what is the null for the f test of weak instruments?
instruments in the first stage are weak
what does it mean if you reject f test for weak instruments?
instruments explain variation in endogenous variables–
what does it mean if you fail to reject the null of the test for weak instruments?
You need to find a betterr instrument
what is the general rule of thumb for test of weak instruments (value greater than __)?
10–if it’s greater than 10, it is not a weak instrument
How do you use f test in context of RD?
to see if there are other points of discontinuity
what is the null of the f test in context of rd (checking for other points of discontinuity)?
There are not other points of discontinuity–dummies at other percentiles are not stat sig
what does it mean if you reject the f test in context of rd (checking for other points of discontinuity?)
there are other points of discontinuity–the relationship between x and y is not smooth
in propensity score matching, for the balance test, what is the null of the t test?
the means of the two groups (T and C) are equal
what does it mean if you reject null for balance test?
means of the two groups are no equal
how do you test for manipulation in RD?
mccrary test-is there clumping?
what tests can you use to test for heteroskedasticity?
Breush-Pagan, White
what is the null of the breusch-pagan test
model is homoskedastic
what does it mean if you reject the null of breusch-pagan?
you have heteroskedasticity–use robust standard errors!
how do you calculate the variance inflation factor?
1/(1-R^2j)
what is the measure for multicollinearity?
VIF
If VIF is less than 4, should you be worried about multicollinearity?
no
If VIF is greater than 10, should you be worried about multicollinearity?
yes
