Title | ECO 045 EXAM 3 Study Guide |
---|---|
Author | Ivy Yu |
Course | Statistical Methods |
Institution | Lehigh University |
Pages | 4 |
File Size | 119.2 KB |
File Type | |
Total Downloads | 4 |
Total Views | 32 |
ECO 045 EXAM 3 STUDY GUIDE5. Hypothesis Tests ● Hypothesis Test - formal procedure to answer a specificquestion about the value of population parameters ○ Possible errors in test results 𝐻𝑜 Accept𝐻𝑜 Reject𝐻𝑜True Correct Type I ErrorFalse Type II Error Correct ■ Type I Error - null hypothesis is true...
ECO 045 EXAM 3 STUDY GUIDE 5. Hypothesis Tests ● Hypothesis Test - formal procedure to answer a specific question about the value of population parameters ○ Possible errors in test results 𝐻
Accept 𝐻
Reject 𝐻
True
Correct
Type I Error
False
Type II Error
Correct
𝑜
● ●
𝑜
■ Type I Error - null hypothesis is true, but we reject it ■ Type II Error - null hypothesis is false, but we do not reject it ■ Otherwise the test conclusion is correct Null Hypothesis (Ho) - a default assumption, typically being challenged with the analysis Alternative Hypothesis (Ha) - the logical opposite of the null hypothesis ○ Null Hypothesis and One-Sided Alternatives: ■ 𝐻 :µ ≥ µ , 𝐻 :µ < µ 0
■ ○
0
0
0
0
𝑎
0
Null Hypothesis and Two-Sided Alternatives: ■ 𝐻 :µ = µ ,𝐻 :µ ≠ µ 0
𝑎
0
Test Statistic - the statistic whose value is used to determine the outcome of the hypothesis test Level of significance ○
●
𝑎
𝐻 :µ ≤ µ ,𝐻 :µ > µ
0
● ●
𝑜
Probability of making Type I Error: 𝝰
■ Favors null hypothesis as default assumption ○ Probability of making Type II Error ■ Power of test: 𝞫 Critical Values - threshold in test statistic that is used to determine whether 𝐻 should be rejected 𝑜
●
●
Find Critical Values in the z-score: ○
One-sided alternative, where alternative is below: F(z*) = α
○
One-sided alternative, where alternative is above: 1 - F(z*) = α
○
Two-sided alternative: F(-z*) =
α 2
Find Critical Values in the t-score (t-distribution): (*use when the sample size is below 100 and σ is unknown*) ●
One-sided alternative, alternative is above null: ○ Table shows upper tail probability, P(...> t) ○
●
α
One-sided alternative, alternative is below null: ○ t-distribution is symmetric, P(...> t) = P(...< -t) ○
●
Find entry where DF=n-1 and probability =
Find entry where DF=n-1 and probability =
○ Use -t to find critical value Two-sided alternative:
α
○ ●
α
Find entry DF=n-1 and probability = 2
z-score for Hypothesis Test:
=
𝑥−µ0
○
z-score for Sample Mean: 𝑧
○
t-score for Sample Mean (use if n < 100 and σ is unknown: 𝑡 z-score for Sample Proportion:𝑧
○
σ𝑥
=
z-score for Difference in Means: 𝑧
○
=
𝑥−µ0 𝑠/ 𝑥
𝑝−𝑝0
=
σ𝑝 𝑥1 − 𝑥2 σ𝑥 − 𝑥 1
2 2
■
Standard Error of Difference in Means: σ
𝑥1− 𝑥2
=
𝑉(𝑥1 ) + 𝑉(𝑥2) =
Critical Value Approach 1. State null and alternative hypothesis 2. Specify level of significance (α = 0.01, 0.05, 0.10) 3. Designate test statistic (and data source) a. We focus on using the z-score (or t-score) as the test statistic 4. Determine critical values 5. Compute test statistic 6. Compare with critical values and conclude a. Reject 𝐻 : 0
i. z < -z* ii. z > +z* b. Do not reject 𝐻 (Accept 𝐻 ): 0
0
i. -z* < z < +z* p-Value Approach 1. State null and alternative hypothesis 2. Specify level of significance (α = 0.01, 0.05, 0.10) 3. Designate test statistic (and data source) a. We focus on using the z-score (or t-score) as the test statistic 4. Compute test statistic 5. Compute p-Value of test statistic 6. Compare with α and conclude a. Reject 𝐻 : p < α 0
b. Do not reject 𝐻 (Accept 𝐻 ): p ≥α 0
0
Goodness-of-Fit Test 1. State null and alternative hypothesis eg. 𝐻 : 𝑃 = 𝑥 , 𝑃 = 𝑦 , 𝑃 = 𝑧, etc. 0
𝐴
𝐵
𝐶
𝐻 : 𝑃 ≠ 𝑥, or 𝑃 ≠ 𝑦, or 𝑃 ≠ 𝑧,or etc. 𝑎
𝐴
𝐵
𝐶
σ1 𝑛
1
+
σ 𝑛
2 2
2
2. Specify level of significance (α = 0.01, 0.05, 0.10) (𝑓 −𝑒 )
2
𝑖
3. Designate test statistic: Chi-square distribution: χ = ∑
2
𝑖
𝑒
𝑖
𝑖
4. Determine Critical Value from Chi-square table (DF → row, α→ column) 5. Compute test statistic 6. Compare and conclude 2
2*
a. Reject 𝐻0: χ > χ
2
2*
b. Do not reject 𝐻0(Accept 𝐻0): χ ≤ χ 6. Regression Analysis ●
Sample Covariance: 𝑠
●
Correlation Coefficient:
𝑆𝑥𝑦
𝑟𝑥𝑦 =
𝑆𝑥𝑆𝑦
𝑥𝑦
=
Σ(𝑥𝑖−𝑥)(𝑦𝑖−𝑦) 𝑛−1 Σ(𝑥𝑖−𝑥)(𝑦𝑖−𝑦) 𝑛−1
= (
1 𝑛−1
2
1 𝑛−1
Σ(𝑥 −𝑥) ) 𝑖
2
2
𝑖
Regression model (population): 𝑦 = β + β 𝑥 + ϵ
●
Population regression line: 𝐸(𝑦|𝑥) = β + β 𝑥
2
2
2
[𝑛Σ𝑥 −(Σ𝑥) [𝑛Σ𝑦 −(Σ𝑦) ]
Σ(𝑦 −𝑦) )
●
0
𝑛(Σ𝑥𝑦)−(Σ𝑥)(Σ𝑦)
=
1
0
1
○ 𝐸(𝑦|𝑥) : expected value (mean) of y given x ○ β , β Parameters: 0
1
■
β : slope 1
■ β : intercept 0
● ●
Mean of 𝑦 if 𝑥 = 0
Sample regression line: 𝑦 = 𝑏 + 𝑏 𝑥 0
1
○ 𝑦 : predicted value of y given x ○ 𝑏 , 𝑏 : Parameter estimates 0
1
■ 𝑏 (slope estimate): 𝑏 1 1
=
Σ(𝑥𝑖−𝑥)(𝑦 −𝑦) 𝑖
2
Σ(𝑥 −𝑥)
𝑠𝑥𝑦
=
𝑠
2
𝑥
𝑖
■ 𝑏 (intercept estimate): 𝑏 = 𝑦 − 𝑏 𝑥 0
●
0
1
Sums of Squares: 2
○
Total Sum of Squares: TSS= Σ(𝑦𝑖 − 𝑦)
○
Sum of Squared Errors: SSE=Σ(𝑦𝑖 − 𝑦𝑖)
○
Sum of Squares due to Regression: SSR= Σ(𝑦𝑖 − 𝑦)
○
TSS = SSR + SSE
2 2
2
𝑆𝑆𝑅 𝑇𝑆𝑆
= 1−
𝑆𝑆𝐸 𝑇𝑆𝑆
2
= (𝑟𝑥𝑦)
●
Coefficient of Determination (measure of fit): 𝑟 =
●
Testing for the significance of β : Can use critical value or p-value approach 1
“statistical significance” - statistical evidence that a parameter is nonzero at significance level α ○ Hypotheses: 𝐻0:β1 = 0 𝐻𝑎:β1 ≠ 0
○
z-score for Significance Test (test statistic): 𝑧
■
Standard error of 𝑏 (𝑠 ): 1
●
𝑏1
𝑏1 𝑠
=
𝑏1
𝑠
2
ϵ
𝑠𝑏 =
2
Σ(𝑥𝑖−𝑥)
1
depends on the size of error terms vs. sample size and amount of variation in x 2
●
2
𝑠ϵ =
Σ(𝑦𝑖−𝑦𝑖)
𝑆𝑆𝐸 𝑛−2
=
𝑛−2
2
2
● Σ(𝑥𝑖 − 𝑥) = (𝑛 − 1)𝑠𝑥 ■
Sampling distribution of 𝑏1: 𝑏
1
●
●
=
𝑠𝑥𝑦 2
𝑠𝑥
2
𝑏 ~ 𝑁𝑜𝑟𝑚𝑎𝑙 (β , 𝑠 ) 1
1
𝑏1
○
Mean: β
○
Variance: 𝑠
1 2
𝑏1
Key Assumptions for Regression Analysis ○ Error terms ϵ are independent of 𝑥 ○ There is variation in 𝑥: 𝑉(𝑥) > 0 ○
2
In small samples (n...