1 / 39

Hypothesis Testing: One Sample Mean or Proportion

Hypothesis Testing: One Sample Mean or Proportion. P - Values, Sections Section 10.3 Testing a Proportion, Section 10.6 Power of a Test, Section 10.7. P-value Approach. State null and alternative hypotheses H 0 :  = 8000 H 1 :   8000 (two sided test)

roshaun
Download Presentation

Hypothesis Testing: One Sample Mean or Proportion

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: One Sample Mean or Proportion P - Values, Sections Section 10.3 Testing a Proportion, Section 10.6 Power of a Test, Section 10.7

  2. P-value Approach • State null and alternative hypotheses • H0:  = 8000 • H1:  8000 (two sided test) • Set level of significance, ⍺ = 0.05 • Do not find the critical values • Calculate appropriate test statistic • Z = (7814.1-8000)/94.28 = -1.97 PP 3

  3. P-value • The p-value tells us the probability of observing a test statistic as extreme or more extreme than the one we observed if the null hypothesis is true • In this example, the p-value tells us the probability of observing the test statistic of -1.97 or one more extreme if  = 8000 • Think more extreme as |1.97| PP 3

  4. Sampling Distribution of Sample Means, Normally Distributed .4756 8185.5 7814.1 -1.97 0 1.97 What is the p-value of |1.97|? • Need P(0 < z < 1.97) = 0.4756 • Standard normal table • In one tail = .5000 - .4756 = .0244 • In two tails = .0244*2 = .0488 • p-value = .0488 • This tells us that if the null hypothesis is true we would observe a test statistic as extreme as 1.97 in 4.88 samples out of 100 samples (4.88% of the time) .0244 .0244 Z PP 3

  5. Making the Statistical Decision • Decide whether the p-value is small and implies the sample mean is a rare event if the null hypothesis is true or whether the p-value is large and implies the sample mean could be expected to occur if the null hypothesis is true • Use our conventional levels of alpha (levels of significance) as benchmarks • Compare the p-value we calculate from the data with the level of significance PP 3

  6. Compare the p-value with ⍺ • If  = .05, then • If p-value < .05, reject the null hypothesis • If p-value ≥ .05, do not reject the null hypothesis • If  = .01, then • If p-value < .01, reject the null hypothesis • If p-value ≥ .01, do not reject the null hypothesis PP 3

  7. P-value and Critical Values - Consistent • If the test statistic (|1.97|) is greater than the critical value (|1.96|), the p-value (.0488) must be smaller than .05 • We will reject the null hypothesis whether we use a critical value or a p-value approach PP 3

  8. Observed Significance Level • A p-value is: • the exact level of significance of the test statistic • the smallest value a can be and still allow us to reject the null hypothesis • the amount of area left in the tail beyond the test statistic for a one-tailed hypothesis test or • twice the amount of area left in the tail beyond the test statistic for a two-tailed test. • the probability of getting a test statistic from another sample that is at least as far from the hypothesized mean as this sample statistic is PP 3

  9. One Tail Tests • The researcher is interested in whether the population parameter is lower (or higher) than the stated value in the null • Candy manufacturer claims there are 100 candies on average in each $1.00 bag • After sampling a number of bags, you are concerned that this claim may be an overstatement • Do we care if there are more than 100 candies in a bag on average? PP 3

  10. Example of Lower Tail Test • Form the alternative hypothesis first, since it embodies the challenge • H0: μ = 100 or H0: μ ≥ 100 (No evidence to reject the manufacturer’s claim) • The equality is always in the null • H1: μ < 100 (The manufacturer’s claim is rejected; Mean number of candies is less than 100) PP 3

  11. Lower Tail Test • The rejection region is in the lower tail of the distribution • All of ⍺ is placed in the one tail • Reject the null • If we observe a small sample mean that is unlikely to occur if μ = 100 Sampling Distribution under the null hypothesis reject Do not reject PP 3

  12. Example of Upper Tail Test • A company is thinking about setting up an on-site day-care program for its employees. The CEO has stated that she will do so only if more than 80% of the employees favor such a decision. • When will an action take place? • H0: π ≤ 0.80 • H1: π > 0.80 PP 3

  13. Two Sided Tests • Always use = and ≠ in the statistical hypotheses • Directionless in that the alternative hypothesis allows for either the greater than (>) or less than (<) possibility PP 3

  14. One tailed tests • Are always directional • The alternative hypothesis uses either the greater than (>) or less than (<) sign • Only used when the researcher • Knows for certain that the outcome of an experiment is going to occur only in one direction • Is only interested in one direction of the experiment • In one-tailed problems, the researcher is trying to demonstrate that something is • Older, younger, higher, lower, more, less, greater, less than, has increased, has decreased and so on • The directionality to be demonstrated is placed in the alternative hypothesis PP 3

  15. Problem – Lifetime of Light Bulbs • The advertising campaign of a light bulb manufacturer claims that the mean lifetime of their bulbs is 1600 hours • The known population standard deviation is  = 120 hours • In your job for Consumer Report you have been asked to check whether the company is making false claims • You obtain a sample of 100 bulbs • The mean lifetime of the sample is 1,570 hours • Let  = .05 • Sample means are normally distributed via the CLT PP 3

  16. Problem – Lifetime of Light Bulbs • H0: ≥ 1600 • The company’s claim appears valid • H1:  < 1600 • The company’s claim appears to be an overstatement • Find critical value Z.05 = -1.64 • Look for probability of .4500 in standard normal table • Look for probability of .05 in t table for infinite df • DR: if (Z test statistic ≥ -1.64) Do not reject H0 • Z = (1570 - 1600)/(120/100) = -2.50 PP 3

  17. Sampling Distributionof the sample mean under the null hypothesis ⍺ = 0.05 Do not reject 1570 1600 -2.50 -1.64 Z Lifetime of Light Bulbs: Critical Value Approach • DR: if (Z test statistic ≥ -1.64) do reject H0 • Calculate test statistic • Z = (1570 - 1600)/(120/100) = -2.50 • Reject the H0 • The company appears to be overstating the lifetime of its bulbs -1.64 PP 3

  18. Lifetime of Light Bulbs: p-value Approach • H0: ≥ 1600 (The company’s claim appears valid) • H1:  < 1600 (The company’s claim appears to be an overstatement) • Sample means are normally distributed via the CLT • Let ⍺ = 0.05 • Find test statistic: • Z = (1570 - 1600)/(120/100) = -2.50 PP 3

  19. -2.50 0 Lifetime of Light Bulbs: p-Value Approach • How likely are we to observe a test statistic of -2.50 or one more extreme, if the null hypothesis is true? • Standard Normal Table • P(0 < Z < 2.50) =.4938 • Calculate area in tail • .5000 - .4938 = .0062 • Compare with level of significance • .0062 < .05, reject null Sampling Distributionof the sample mean under the null hypothesis p-value = .0062 .4938 1570 1600 Z PP 3

  20. The t-Test Statistic • Consider the population standard deviation, , is unknown • Sampling distribution of Sample Means is normal because • Population follows the normal distribution • Or the sample size is greater than 30 (CLT) • How do we estimate the standard error of the sample means? • Substitute S, the sample standard deviation, for  PP 3

  21. T-test Problem: Soft Drink • A soft drink manufacturer wants to test whether the mean rating for a new flavor that it has just developed equals 60. From many previous studies it is known that its old flavor has a mean rating of 60. • If the mean rating for the new flavor is higher than 60, the manufacturer will substitute the new flavor for the old and stop producing the latter. • The manufacturer samples 16 people and obtains ratings on a scale of 1 to 90. The sample mean is 62.38 and the sample standard deviation is 7.6. • Should the manufacturer continue to produce the old flavor or should it shift production to the new flavor? • The ratings are normally distributed in the population. Let  = .05 PP 3

  22. T-test Problem • Where is the action? • If the mean rating for the new flavor is higher than 60, the manufacturer will substitute the new flavor for the old and stop producing the latter • H0: μ ≤ 60 (The new is the same or worse than the old. Do not switch) • H1: μ > 60 (The new is better than the old. Switch) • One sided, upper tail test PP 3

  23. Sampling Distributionof the sample mean under the null hypothesis 62.38 1.25 1.753 T-test Problem • Estimate the standard error • Calculate the test statistict15 = (62.38 - 60)/1.9 = 1.25 • Find the CV – t table • t15,.05 = 1.753 • Do not reject • Do not switch to new reject ⍺ = .05 Do not reject 60 Z PP 3

  24. Testing Hypotheses about the Population Proportion • Proportions are used for qualitative data • Observations fall into one category or another • Will you vote for the candidate: yes or no • Do you like the flavor: yes or no • We are interested in some characteristic of the observation • Population proportion is of interest • π = number of observations with characteristic/ population size = x/N PP 3

  25. Testing Hypotheses about the Population Proportion • Have the sample information • p = observations with characteristic/ sample size = x/n • If we had drawn a different sample, what might p in that sample be? • Need to understand the sampling distribution of the sample proportion • Can calculate exactly what the sampling distribution looks like PP 3

  26. The Sampling Distribution of the Sample Proportion • Suppose I have a large population • Fifty percent vote yes on an issue and fifty percent vote no • Interested in the proportion voting yes (a success) • Let π = 0.50 • Draw a sample of n = 5 • What would the distribution of sample proportions look like? • Answer lies with the binomial distribution • “What is the probability of observing exactly 0 (1, 2, 3, 4, 5) “successes” in 5 trials if the probability of success on a single trial is 0.5? PP 3

  27. The Binomial Distribution Use the Binomial Distribution tables in the appendix of Weiers to find the probability of observing exactly 0 successes in 5 trials, or 1 success in 5 trials and so forth. x = number of successes Proportion of successes Probability of x successes Using the binomial distribution, we can calculate the probability of observing certain sample proportions, given the population proportion is some specific value π = 0.50 and given the sample size, n = 5. PP 3

  28. Sampling Distribution of the sample proportion, p, when n = 5 and π = 0.50 P(x/n=p) = p The Sampling Distribution of the Sample Proportion • Is a binomial distribution • Not a continuous distribution • E(x/n) = E(p) = π • VAR(p) = π(1 - π)/n PP 3

  29. E(p) = π p Normal Distribution as an Approximation to the Binomial Sampling Distribution of p when n is “large” • Sampling distribution of the proportion will be approximately normal with mean π and standard deviation p • Conditions to be met • nπ > 5 when π 1/2 and n(1 - π) > 1/2 when π > 1/2 • Or sample size greater than 100 PP 3

  30. Hypotheses about the Population Proportion • Formulate the hypotheses • H0: π = π0 • H1: π ≠ π0 • or H1: π > π0 or H1: π < π0 • Set ⍺ and look-up Z critical value(s) in table • Form the rejection region assuming the null hypothesis is correct • Or set ⍺ and use the p-value approach PP 3

  31. Reject Reject E(p) = π0 p -Z⍺/2 Z⍺/2 Z Standard Error = Hypothesis about the Population Proportion Assuming the null hypothesis is correct, two sided test • Calculate the test statistic • Use π0in the standard error • If the test statistic is extreme, reject the H0 • DR: if (-Z⍺/2 ≤ Z test statistic ≥ Z⍺/2 ) do not reject H0 PP 3

  32. Proportion Problem • The manufacturer of an over-the-counter medicine claims that it is 90% effective in relieving an allergy for a period of 8 hours • In a sample of 200 persons who have the allergy, the medicine provided relief for 160 people • Determine whether the manufacturer’s claim is legitimate (overstating value?) • Let  = .01 PP 3

  33. Proportion Problem - OTC Med • Characteristic of interest • Obtaining relief • H0: π ≥ 0.90 • The claim is correct and any observed sample difference is due to chance • H1: π < 0.90 • The claim is false and the observed sample proportion is unlikely to have come from a population with a π of 0.90 PP 3

  34. Reject p Proportion Problem - OTC Med Sampling Distribution of p • Find critical value • Z.01 = -2.33 • DR: if (Z test statistic ≥ -2.33) Do not reject H0 • What do we observe in the sample? • p = x/n = 160/200 = 0.80 • Create test statistic • Z = (.80 - .90)/ 0212 = -4.71 • Statistical decision • Reject the H0 • Conclusions • The claim is not legitimate p PP 3

  35. Look at t- table PP 3

  36. .80 .90 p - 4.71 0 Z p - Value Approach • How likely are we to observe a sample proportion of 0.80 or one more extreme if the population proportion is 0.90? • Standard normal table • P(0 < Z < 4.71) ≈ 0.50 • There is ≈ .000 in the tail • p - value = .000 • p - value < .01, Reject .000 ≈ 0.50 PP 3

  37. Standard Normal PP 3

  38. Weier’s Z Tables ≈ 0.500 ≈ .000 - 3.10 0 - 3.10 Z Any Z value beyond 3.09 can be viewed as a percentile point that contains roughly 50% of Z values between it and 0 PP 3

  39. Online Homework - Chapter 10 Overview of Hypothesis Testing • CengageNOW fourth assignment PP 3

More Related