1 / 25

Hypothesis Testing

Hypothesis Testing. Ch 10, Principle of Biostatistics Pagano & Gauvreau Prepared by Yu-Fen Li. Statistical Inference. Estimation of parameters point estimation interval estimation Tests of statistical hypotheses construct a confidence interval for the parameter

yanni
Télécharger la présentation

Hypothesis Testing

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. Hypothesis Testing Ch 10, Principle of Biostatistics Pagano & Gauvreau Prepared by Yu-Fen Li

  2. Statistical Inference • Estimation of parameters • point estimation • interval estimation • Tests of statistical hypotheses • construct a confidence interval for the parameter • conduct a statistical hypothesis test • critical value • p-value

  3. Steps of hypothesis testing • Predetermine a significance level (α) • Usually set at 0.05 • State the null (H0) & alternative (Ha) hypotheses • Collect sample data • Calculate the test statistic • Construct the CI / find the critical value or p-value • Make the conclusion • reject H0 / do not reject H0

  4. Ho and Ha • Ho and Ha are mutually exclusive and exhaustive (i.e. one of the two statements must be true) • Two-sided tests • One-sided tests • v.s. • A two-sided test is always the more conservative choice • the p-value of a two-sided test is twice as large as that of a one-sided test

  5. P-value • Given that Ho is true, the probability of obtaining a mean as extreme as or more extreme than the observed sample mean (or test statistic) is called the p-value of the test • If p is less than or equal to α, we reject H0 • If p is greater than α, we do not reject H0 Ho

  6. Reject or not reject • a mathematical equivalence between CIs and tests of hypothesis • Any value of z that is between -1.96 and 1.96 would result in a p-value greater than 0.05. • H0 would be rejected for any value of z that is either less than -1.96 or greater than 1.96. • At the α = 0.05 level, the numbers -1.96 and 1.96 are called the critical values of the test statistic.

  7.  Common Misunderstandings of Statistical Hypothesis Testing • The p value is the probability that the null hypothesis is incorrect. • WRONG! The p value is the probability of the current data or data that is more extreme assuming H0 is true • Small p values indicate large effects. • WRONG! p values tell you next to nothing about the size of an effect

  8. Z-Tests and t-Tests • Assume that the continuous r.v. X has mean μ0 and the known standard deviation σ • When n is large enough, the test statistic (TS) • When the standard deviation σ is not known, we substitute the sample value s for σ. • If X is normally distributed, the test statistic (TS)

  9. Example 1: Z-test • A random sample of 12 hypertensive smokers has mean serum cholesterol level = 217 mg/100 ml. Is it likely that this sample comes from a population with mean = 211 mg/100 ml? the area to the right of z = 0.45 is 0.326 Therefore, the p-value of the test is 0.652. If the significance level is set to be α = 0.05, we do not reject Ho for p > 0.05

  10. Example 2: t-test • The underlying distribution of plasma aluminum levels for this population is approximately normal with unknown mean μ and standard deviation σ. Consider a random sample of 10 children selected from a population of infants receiving antacids containing aluminum with mean plasma aluminum levels 37.20 μg/l and standard deviation 7.13 μg/l. • Is it likely that the data in our sample could have come from a population of infants not receiving antacids with mean μ0 = 4.13 μg/l? the total area to the right of 14.67 and to the left of -14.67 is less than 2(0.0005)=0.001. Therefore, p < 0.05, and we reject the null hypothesis

  11. Example: one-sided test • Consider the distribution of hemoglobin levels for the population of children under the age of 6 who have been exposed to high levels of lead. • we are concerned only with deviations from the mean that are below μ0=12.29 g/100 ml, the mean hemoglobin level of the general population of children: vs

  12. Example: one-sided test • A random sample of 74 children who have been exposed to high levels of lead has a mean hemoglobin level of 10.6 g/100 ml with σ = 0.85 g/100 ml. the area to the left of z is less than 0.001. Since this p-value<α = 0.05, we reject Ho. Note 12.29 lies above 10.8 which is the upper one-sided 95% confidence bound for μ in pervious chapter

  13. A slide in Ch 9 Example: an one-sided confidence interval • A sample of 74 kids who have been exposed to high levels of lead from a population with an unknown mean μ and standard deviation σ = 0.85 g/100 ml • These 74 children have a mean hemoglobin level 10.6 g/100 ml. Based on this sample, a one-sided 95% CI for μ - the upper bound only – is • We are 95% confident that the true mean hemoglobin level for this population of children is at most 10.8 g/100 ml

  14. Types of Error • Two kinds of errors can be made when we conduct a test of hypothesis Ho is true Ha is true

  15. Example • The mean serum cholesteric levels for all 20- to 74-yr-old males in US is μ = 211 mg/100 ml and the standard deviation is σ = 46 mg/100 ml. • If we do not know the true population mean but we know the mean serum cholesterol levels for the subpopulation of 20- to 24-yr-old males is 180 mg/100 ml • What is the probability of the type II error associated with such a test, assuming that we select a sample of size 25 and at the α = 0.05 level of significance?

  16. Example: Step 1 Find the cutoff under the Type I error rate • find the mean serum cholesterol level for Ho to be rejected for α = 0.05 (pink area) Ho

  17. Example: Step 2 Find the type II error rate • What is the chance of obtaining a sample mean that is less than 195.1 mg/100 ml given that the true population mean is 211 mg/100 ml? Ha

  18. Comment • The type II error is calculated for a single such value, μ1. • If we had chosen a different alternative population mean, we would have computed a different value for β. • The closer μ1 is to μ0, the more difficult it is to reject the null hypothesis. • β (yellow area) is bigger when μ1 is closer to μ0

  19. Statistical Power • The power of the test of hypothesis is the probability of rejecting the null hypothesis when H0 is false, which is 1 - β. • In other words, it is the probability of avoiding a type II error

  20. Power Curve

  21. α ↑ but β ↓, vice versa

  22. Sample Size Estimation • If we conduct a one-sided test of the null hypothesis • If we conduct a two-sided test, the sample size necessary to achieve a power of 1 - β at the α level is

  23. Under H0, • Under Ha, c

  24. Sample size calculation • A one-sided test will be conducted at the  = 0.05. • Assume =0.85, 0=11.79, and 1 =12.29 • We want a power of 0.80 (1-) • Therefore, • A sample of size 18 would be required

  25. What are the factors influencing the statistical power of a test? •  • n •  •  • |0-1|

More Related