Stat 217 – Day 19RemindersLast Time: Test of Significance for a Population Mean (p. 395, 402)Activity 20-1 (p. 397)Other explanationsLast Time – One sample t-interval (p. 375, 384)Last Time – One Sample t IntervalPreviously- Statistical InferencePreviously – Tests of SignificancePreviously - Confidence IntervalsData CollectionHandoutLab 3: JFK vs. JFKCSolution approach 1Solution Approach 2: Central Limit TheoremButSolution Approach 2Chocolate chip vs butterscotchBUTSlide 20Stat 217 – Day 19Two-sample proceduresRemindersAdditional examples in Blackboard and Self-check ActivitiesLast Time: Test of Significance for a Population Mean (p. 395, 402)1. Let represent a population mean2. State competing hypotheses about Ho: = 0Ha: <,>, ≠ 03. Check technical conditions- Random sample from population of interest- Large sample or normal population (check by looking at the sample)4. Test statistic5. p-value (computer)6. Conclusion about Ho, summarize in Englishnsxt/0Activity 20-1 (p. 397)With a p-value of .0023, we have strong evidence to reject the null hypothesis(n) Because p-value is small, we would reject at the .10 level, the .05 level, the .01 level, the .005 level!Pretty surprising for a random sample of 25 games from a population with =183.2 to give a sample mean of 195.88 or larger.About .23% of random samples from a population with = 183.2 will have a sample mean of 198.55 or largerHave eliminated “random chance” as a plausible explanation for the difference we observedOther explanationsSample size was not especially large, our p-value calculation could be off (CLT)But sample was reasonably normalNot a random sample from entire seasonGames early in the season may be higher scoring than later in seasonNot an experimentEven if decide it’s convincing that scoring has gone up on average, can’t say it was because of the rule changesLast Time – One sample t-interval (p. 375, 384)We are 95% confident that (after the rule change) the population mean points scored per game () is between 187.5 and 204.25Meaning 95% of all intervals constructed this way will succeed in capturing Last Time – One Sample t IntervalNot a prediction interval (p. 381)Effects of confidence level, sample size, and now sample variability (s) as wellLess variability between samples gives “better” (more precise) estimate of population meanLess variability between observational units gives “better” (more precise) estimate of population meanPreviously- Statistical InferenceConfidence Intervals: Goal is to estimate population parameter based on sample statisticInterval of believable values for parameterTests of Significance: Goal is to assess how much evidence have against a particular claim about the parameter based on sample statisticp-value => strength of evidence against H0Previously – Tests of SignificancePopulation proportion H0: Technical conditionsSRS Test statisticp-value: computerPopulation mean H0: Technical conditionsSRSn>30 or normal populationTest statisticp-value: computer10)1(,1000nnnpz/)1(ˆ000nsxt/0 % of samples with statistic at least this extreme when Ho truePreviously - Confidence IntervalsPopulation proportion Common z* values95%: z* 2 Technical conditionsSRS Population mean t* > z*Technical conditionsSRSn>30 or normal population (look at sample))/(*nstx nppzp /)ˆ1(ˆ*ˆ10)ˆ1(,10ˆ pnpnData CollectionFlip a coinHeads = chocolate chipTails = butterscotchPlace the chip on the top of your tongue and hold it to the roof of your mouth and record how long it takes to melt (in seconds)StopwatchHandout(a) How would we decide whether this is convincing evidence that the average melting time among all Cal Poly students differs from one minute?one-sample t-test(b) How would we decide whether one type of chip takes longer to melt, on average?Lab 3: JFK vs. JFKC1. Define parameter of interest2. State Ho and Ha about parameterSolution approach 1“Randomization test” (Lab 3)Difference in means ( )21xx Should center around zero (Null hypothesis)Rather symmetric shape!7.29 pretty surprisingSolution Approach 2: Central Limit TheoremThe randomization/sampling distribution of the difference in sample means will be 1. approximately normal2. mean equal to 1-23. standard deviation equal toas long asrandom assignment or random samplesNormal populations or large sample sizes222121nnButWill instead useAnd compare to the t distributionTest of Significance Calculator: Two means222121nsnsSolution Approach 2Chocolate chip vs butterscotch1. Define parameter of interest2. State Ho and Ha about parameter3. Check the technical conditions4. Calculate the test statistic5. Calculate the p-value6. Draw conclusions (Reject Ho? Answer research question!)If given a “level of significance ,” can use that as the cut-off to determine whether your p-value is “small”BUTIs there a better way to compare chocolate chip vs. butterscotch chip melting times?(is there a better way to decide whether strength shoes improve jumping ability)Why better?Key: Getting rid of a (uninteresting) source of variability…To turn in with partner:(h) Interpret the p-value (what is it the probability of?). Then state your final decision in context.(i) Interpret the confidence interval (Hint: Remember what our parameter is)For ThursdaySelf-check Activity 22-4 (not part d)No pre-labBy MondayTopic 23 (23-1, 23-5), Review Lab
View Full Document