Final ReviewFall 2011Time and PlaceUniversity Set.Psychology 113Saturday 12/17/2011. 12:25 PM - 2:25 PMBasicsCalculator. NO laptop, NO cell phones (NO otherelectronic devices).Coverage: CumulativeChapter 1 - Chapter 10 (plus my extra notes on sampling,simulation, etc)Focus on Chapter 7- Chapter 10Not responsible for all of the sections in the book.What is covered in my posted lecture notes and theassigned HW problems.Chapter 13 is a way to practice what you learned.Closed notes, closed book.Two formula sheets 812× 11 sheet (both sides).Other Questions?Practice ProblemsLook at the HW ProblemsLook at the HW Solutions - gives ideal for getting partialcredit.Examples covered in lecture notes.Examples in your textbook.Additional practice problems were emailed ( withsolutions).Tables to KnowStandard Normal Distributiont Distributionχ2distributionMidterm 1 Material (Chapter 1 - 6)See the posted midterm review.Analysis of two samples - Chapter 7 and Chapter 9From the design of the experiment:Determine whether you deal with independent samples orpaired samples.Determine from the text whether a prior knowledgeenables you to do a one-sided (directional) test.Detect indications that the data do not form a randomsample. Presence of blocks?Using the actual data,Perform the independent-samples t-test andpaired-samples t-test. Calculate confidence intervals forthe difference in two population means.Determine whether a t-test is a valid method for a givendata set (normal probability plots, sample sizes?).The sign test.Analysis of two samplesIn all tests:The basic ideas of hypothesis tests.Compute a p-valueMeaning of a p-valueType I and II errorsChapter 8. Experimental designObservational versus experimental designsConfounded effectsRead a description of the experiment and identify potentialproblems (confounding variable or lack of control group,...)Chapter 10. Analysis of categorical data (counts)From the text:Recognize and formalize the claim. Goodness-of-fit orindependence question?Using the actual data - contingency tables:Build contingency tables from the text.Perform the chi2tests.Using the actual data - two samples:Relative riskodds ratiodifference between two proportions.Exam FormatExam FormatNot like a single HW problemEach part of the question will be like a HW problem.Maybe one or two problems will have a “stretch” part. Youhaven’t seen it in HW but we have discussed it in class.Potential QuestionsOne question from Chapter 10 categoricalOne question from Chapter 8 experimental design. Verbaldescription of experiments - identify observational vs.experimental. Identify potential problems with theexperiment (e.g. confounding factors), etc.One combined question (Chapters 7,9,10,11,12). I giveyou a description of data and you tell me what methodshould be used. Probably several parts. Examples are inChapter 13 (also in the practice problems).There is a lot of material in Chapter 7. And it is related toChapter 9. It is less obvious how I will ask you this. Onebig question? Maybe two smaller questions coveringdifferent areas?Chapter 10 Question FormatThe idea is in a single question to test contigency tables,difference in proportions, odds ratio, etc. (other main ideas fromChapter 10 that we covered).Part A. confidence interval for difference in proportions(interpretation)Part B. confidence interval for odds ratio (interpretation)Part C. Contigency tableOther ConceptsRepeated Sampling Interpretation for confidence intervalsThe Null Distribution of a Hypothesis Test.examples: t with 14 degrees of freedom; χ2with 5 dfNull Distribution - more detailsThis is the reference distribution we use to compute ap-value.For example, say we observed a test-statistic of 1.5 is that“too big”? Well we need to compare it to some referencedistribution, i.e. the distribution of the test statisticsassuming the null hypothesis is true. For simplicity, weshorten this to the null distribution or the null distribution ofthe test statistics.Null Distribution - (continued)(Two sample problems). If we are willing to assumenormality of the underlying populations (or approximatenormality because of the central limit theorem), then thenull distribution is based on the t-distribution, hence werefer to it as a t-test.If we are not willing to assume normality, another option isto use a randomization test. Here, we use a computer toapproximate the null distribution. There are other tests(besides the randomization test) when you are not willingto assume normality (e.g. the sign
View Full Document