ECON 164
Lecture 13
Hypothesis testing, III
p-values: alternative to t-stats
p-value means probability value
Your t-statistic and your p-value are both indicators of the
likelihood of finding what you found if the null were true
If two-sided test, probabi
ECON 164
Lecture 20
Review of functional forms
Recall: Dummy X variable
A way of including a non-numerical variable in
your regression model.
How many values can a dummy variable take?
2: Think of it as an On/Off switch
For a given categorical variable, w
ECON 164
Lecture 17
Specification: Functional
Form
Sensitivity analysis
Most empirical studies report more than one
regression
When you deliberately run (and report) a set of
alternative specifications, to show that your
main estimates are robust
You like
ECON 164
Lecture 16
Specification: Choosing Explanators, III
Menu for today
Omitted vs irrelevant variables
4 specification criteria
Misusing the criteria: An example
Searching for the best specification
Sensitivity analysis
Data mining
Effects of omitted
ECON 164
Lecture 19
Specification: Functional
Form, III
Recall: Quadratic form
Another form that allows the impact on Y of a
change in X to depend on the level of X:
Y = 0 + 1X + 2 X2 +
Advantage over log-log or semilog forms?
Can produce a U-shaped or i
ECON 164
Lecture 15
Specification: Choosing Explanators, II
Recall: Omitted variable bias
When we leave out an explanator, we worry that the
coefficients we estimate will be biased.
Our concern is not the reduced goodness-of-fit, or the
artificially small
ECON 164
Lectures 11 and 12
Hypothesis testing
The null hypothesis: H0
The null hypothesis: A statement about what you
think is not true about the value of
The most common null?
Therefore, what you dont expect to find as your -hat
value
H0: =0 (i.e., tha
ECON 164
Lecture 14
Specification:
Choosing your explanators
Thinking about variables
No chapter devoted to choosing your Y variable
In practice, this is an extremely important choice, can make or
break a project
A viable research project requires a clear
Lecture 2
Econometrics
Overview of regression
analysis
Recall: What constitutes econometrics?
About measuring correlations
Relatively few new statistical concepts
How much do these two variables co-vary?
How confident am I about my estimate?
Sampling dist
ECON 164
Lecture 18
Specification: Functional
Form, II
Recall: Linear in the coefficients
OLS requires only that
f(Y)= 0 + 1f(X)
Recall: Natural logarithm
Natural log mean your base is e
ln X = b means
e is a constant: e=2.71828
b is the natural log of X