Hypothesis Testing & t-Tests Flashcards
What is hypothesis testing?
It is used to determine whether a regression coefficient (π½π) is statistically significant
What are the 2 types of hypothesis?
H0 (null hypothesis): the coefficient is not significant (e.g π½1
= 0)
H1 (alternative hypothesis): the coefficient is significant (e.g π½1 doesnβt = 0)
What is the general structure of hypothesis testing?
1) state the null and alternative hypothesis
2) compute the Test Statistic (e.g t-statistic)
3) compare to a critical value (from the statistical table)
4) make a decision:
- if t is greater than the critical value = reject H0
if t is smaller = fail to reject H0
What is the T test?
Most commonly used test
used to check if an estimated coefficient (π½^π) is significantly different from 0
T test formula
t= Ξ²^jβΞ²0 / SE(Ξ²^j)
βΞ²^j = estimated coefficient
Ξ²0 = hypothesised value (usually 0)
β
SE(Ξ²^j) = standard error of Ξ²^j
Give an example of a T-test
lets use the the following wage regression:
wage =Ξ²0 + Ξ²1 (education) + u
from the OLS we get:
Ξ²1 = 2.5, SE(Ξ²^1) = 0.8
We test:
H0 : Ξ²1 = 0 (education has no effect on wages)
H1 : Ξ²1 =/ 1 (does affect wages)
Now work out the T statstic:
t = (2.5-0)/ 0.8 = 3.125
Compare to the critical value
at the 5% significance level (2 tailed) it is roughly 2
as 3.123<2 we reject H0
what is a two tailed t test used for?
Used when testing whether a coefficient is significantly different from 0, without specifying direction (increase or decrease). The significance level is split across both tails of the t-distribution.
when is a one tailed test used?
Used when you are testing for a specific direction (increase or decrease but not both)
what is a confidence interval?
it is a range of values that we expect the true coefficient (Ξ²j) to be with a certain level of confidence
β
What is the formula for a confidence interval?
Ξ²^jβ +or - (t a/2 x SE(Ξ²^jβ))
Ξ²^jβ = estimated confidence interval
SE(Ξ²^jβ) = standard error of Ξ²^jβ
ta/2 = critical value from the t-table (depends on confidence level)
give an example of using confidence intervals
Letβs say we estimate the following regression model:
wage =π½0+π½1(education)+u
We get the estimated coefficient and standard error:
π½^1=2.5
ππΈ(π½^1)=0.8
Using a 95% confidence interval (π‘πΌ/2=2) we calculate:
2.5Β±(2Γ0.8)
2.5Β±1.6
[0.9,4.1]
This means that we are 95% confident that the true effect of education on wages is between 0.9 and 4.1.
how can we use confidence intervals to test the hypothesis?
We check if zero is inside the confidence interval:
β
If 0 is NOT in the CI (e.g., [0.9, 4.1]) β We reject
π»0, meaning π½1 is statistically significant.
π« If 0 IS inside the CI (e.g., [-1.2, 3.8]) β We fail to reject
π»0, meaning there is not enough evidence that π½1
is different from zero.