1 / 28

Applied Quantitative Methods

Applied Quantitative Methods. Lecture 5 . Statistical Inferences. October 20 th , 2010. Last Lecture Review. Statistical inferences. Estimator (Sample statistic): a function of sample variables TE: sample mean, sample standard deviation Estimate – a realized value of an estimator

mhux
Download Presentation

Applied Quantitative Methods

An Image/Link below is provided (as is) to download presentation Download Policy: Content on the Website is provided to you AS IS for your information and personal use and may not be sold / licensed / shared on other websites without getting consent from its author. Content is provided to you AS IS for your information and personal use only. Download presentation by click this link. While downloading, if for some reason you are not able to download a presentation, the publisher may have deleted the file from their server. During download, if you can't get a presentation, the file might be deleted by the publisher.

E N D

Presentation Transcript


  1. Applied Quantitative Methods Lecture 5. Statistical Inferences October 20th, 2010

  2. Last Lecture Review

  3. Statistical inferences • Estimator (Sample statistic): a function of sample variables • TE: sample mean, sample standard deviation • Estimate – a realized value of an estimator • Sampling distribution: a likelihood of various outcomes of the estimator across different random samples • Statistical inferences • Estimation: providing a range of potential values (confidence interval) for a population parameter based on a sample statistic • - Point & Interval estimators • 2)Testing: using a sample to test the claim about the value of a population parameter

  4. Point Estimation • Method of moments • Equating first k sample moments to the first k population moments and solving the resulting system of equations • X1…..Xn ~iid from a population with f(x|θ1…..θk) • TE Let X1…..Xn ~iid N(θ,σ2)

  5. Point Estimation(Cont.) • Maximum likelihood • For each sample point X, let be a parameter value at which attaints its maximum as a function of θ. • Likelihood function: • - is a random variable • TE: Let X1…..Xn ~iid N(θ,1)

  6. Properties of Estimators • Finite sample properties • Unbiased estimator • The bias is systematic Source: Doane & Seward (2009) Applied Statistics.

  7. Properties of Estimators (Cont.) • Sample mean is an unbiased estimator of the population mean • Sample variance is an unbiased estimator for the population variance • TE: Sample mean vs. First observation in the sample • and • => A need for additional criteria

  8. Properties of Estimators (Cont.) • 2. Efficiency: Unbiasedness + Minimum variance • - Sampling distribution of an estimator should have a small standard error • Sampling variance: • TE: For normal distribution, sample mean and sample variance are MVUEs.

  9. Properties of Estimators (Cont.) • Asymptotic properties of estimators • Consistency: is a consistent estimator of θ if for every ε > 0 • If is consistent, then • A minimal requirement • An unbiased estimator • which • is consistent

  10. Properties of Estimators (Cont.) • 2. Asymptotic normality: shape of the sampling distribution • Central Limit Theorem • Let {X1…..Xn} be a random sample from a population with mean μ and variance σ2, then • The rule of thumb: n ≥ 30 is sufficient to assume normality for a symmetric or slightly skewed population without outliers

  11. Properties of Estimators (Cont.)

  12. Interval Estimation • Confidence interval • Centered on a point estimate • The size of the confidence interval is determined by: • Confidence level (100(1-α) %): more confidence – less precision • Size of σ: larger σ – wider interval • Sample size n: larger n – narrower interval • Methodology of estimation • Case 1: known σ • Case 2: unknown σ

  13. Interval Estimation (Cont.) • Confidence interval for sample mean • CLT: Sampling distribution of follows normal distribution with mean μ and as n increases. • Case 1: known σ • TE: Weekly hours of TV watching • Sample n =25, and a known σ = 10 • Find a 95% confidence interval for the population mean μ • Margin of error: • Critical values (z-scores): • Confidence interval:

  14. Interval Estimation (Cont.) • Case 2: known σ • TE: n=25, and s = 15 • Find a 95% confidence interval for the population mean μ • Margin of error: • Critical values (t-scores): • Confidence interval: • N!B! Interpreting confidence intervals • The probability that μ is in the interval is 95%

  15. Interval Estimation (Cont.) • A 95% confidence interval for the population mean is • - In 95 % of random samples, confidence interval would contain μ

  16. Hypotheses Testing (HT) • Hypothesis is a statement about the value of a population parameter • Hypothesis testing (HT): consistency of data with an assumption about population parameters • - One-sample & Two-sample HT • 1) One-sample HT • The null hypothesis (H0): claim about the value of a population parameter • The alternative hypothesis (H1): • Nondirectional (two-sided): • Directional (one-sided):

  17. Hypotheses Testing (Cont.) • Step-by-step procedure • Formulate H0 and H1 • - mutually exclusive statements • H0 always refers to the population • H1 always refers to a sample • Assume that H0 is true • A benchmark value (prior knowledge, standards)

  18. Hypotheses Testing (Cont.) • 3. Level of significance (α) acceptable level of risk • - α is a probability of Type I error • Conventional levels of significance: 0.01, 0.05 and 0.1 • Confidence level: (1- α)*100% • Type II error (β): • α and β are inversely proportional • To decrease the likelihood of both errors we need to increase the sample size

  19. Hypotheses Testing (Cont.) • 4. Critical values: region of rejection for H0 • For known σ: z-score (standard normal distribution) • For unknown σ: t-distribution • Critical values depend on: • type of the test (one-tailed or two-tailed) • and level of significance (α) • 5. Collect sample • - calculating sample statistics

  20. Hypotheses Testing (Cont.) • 6. Test statistic • Case 1: known σ • Z-score: • Case 2: unknown σ • t-score: • Test statistics measure how far is the value of sample statistic from H0

  21. Hypotheses Testing (Cont.) • 7. Drawing conclusion • Rejection region: the area under the sampling distribution curve that defines an extreme outcome • Left-tailed test: H0 is rejected if zcalc< - zα • Right-tailed test: H0 is rejected if zcalc > zα • Two-tailed test: H0 is rejected if zcalc< - zα/2 or zcalc > zα/2

  22. Hypotheses Testing (Cont.) • One-sample hypothesis test • TE Quality control • Industry standard 216 mm. • Random sample n=50, =216.007, σ=0.023, 5% level of significance • State hypotheses • H0: μ0 = 216 • H1: μ0 > 216 • Level of significance α=0.05 • Critical values of z statistic: • Right-tailed test: z0.95=1.645 • Sample mean 216.007 • Test statistic: • 6. Decision: zcalc falls into rejection region. H0 is rejected at 5 % level of significance

  23. P-value Approach • P-value: What is a probability that we would observe a particular sample mean (or a value even father away from μ0 ), if Ho is true? • - is compared to the level of significance α • Decision rule (right-tailed test): • Reject H0 if P(Z > zcalc)< α, otherwise fail to reject H0 • - the smaller the p-value, the stronger the evidence for rejection H0 • TE zcalc=2.152 • p-value = P(Z > 2. 152) = 1-P(Z< 2.152) = 1-0.9843 = 0.0157 • 0.0157 < 0.05 => H0 is rejected • Interpretation: in a right-tailed test a test statistic of zcalc=2.152 (or a more extreme value) would happen by chance about 1.57 percent of time if H0 is true • Two-tailed test: Reject H0 if 2 P(Z > zcalc)< α, otherwise fail to reject H0

  24. Hypotheses Testing • 2) Two-sample HT • TE Research question: does attendance of lectures boost learning • H0 : There is no difference in midterm test average scores of students who attended lectures and who did not • H0: μ1= μ2 • H1(nondirectional): The average test score in two groups of students will be different • H1:μ1≠ μ2 • Test statistics for three cases: • - both population standard deviations are known • - both population standard deviations are unknown • - both population standard deviations are unknown but equal • Z-score:

  25. Hypotheses Testing Case 1: Margin of error for known σ1 and σ2 : Case 2: Margin of error for unknown σ1 and σ2 :

  26. Next Lecture N!B! Midterm exam: October 27th, at 9:15, RB 212 1. Waldman, M., Nicholson, S. & Adilov, N. (2006). Does Television Cause Autism? NBER Working Paper 12632. 2. Mankiw, Romer and Weil (1992). A Contribution to the Empirics of Economic Growth. The Quarterly Journal of Economics, pp. 407.

More Related