Module 4 and 5 (Confidence, Hypothesis Testing, Selection Bias)) Flashcards
Confidence interval
a range of values where we’re fairly certain that, given some true value, the random variable will fall into
What are the 3 types of confidence intervals?
Centered, right-handed , left-handed
Centered CI
both sides have the same probability (symmetrical)
Right-handed CI
one-sided and only has a lower bound
Left-handed CI
one-sided and only has an upper bound
t-distribution
a “correction” of the normal distribution when n is small
Properties of the t-distribution
symmetrical about the mean, has a mean of 0, approaches normal as n increases
Why are t-based intervals larger than normal-based ones?
they reflect the increased uncertainty caused by using s instead of σ
Hypothesis
a declarative description/statement about something you want to learn about (e.g. a population parameter) that are tested to certain levels of confidence; can be falsified
How is econometrics inductive in nature?
we learn about it by repeated observation and ruling out possible explanations
One-sided hypothesis test
null hypothesis is “less than” or “more than”; a right- or left-handed CI is constructed
Two-sided hypothesis test
null hypothesis is “equal to”; a centered CI is constructed
Size of hypothesis test
alpha or 1 minus the size of the CI
Why would the p-value be small?
the sample size is large, meaning it is easy to statistically detect even small difference (statistical significance)
Type 1 error
false positive; reject the null hypothesis when it is true
Type 2 error
false negative; fail to reject the null hypothesis when it is false
Treatment vs. Outcome
an action which affects a group of people (treated); the impact of the treatment
How is selection bias fundamental?
Cannot be fixed by collecting a bigger sample or by using a more sophisticated comparison
What ways can you deal with selection bias?
use a method which explicitly deal with selection or acknowledge and understand the impact it has on results
Selection bias
arises when the treatment group would have acted differently than the untreated group IF they had been untreated