Stat 217 – Day 14AnnouncementsRecap: Preschool ObesityNew Goals:DefinitionExample: OK City ThunderGraphical summaryConfounding variablesKey IdeaExampleBest Table in Entire CourseSo what next?Stat 217 – Day 14Comparing Two GroupsAnnouncementsNo exams back today Wednesday Office hour moved to noon this week.Pre-lab due by tomorrow morningWatch videoAnswer questions about contextRecap: Preschool ObesityOne research question is how the obesity rates of Caucasians and Hispanics compareIn Lab, looked at “everyone” vs. “Hispanics”But that includes Hispanics in both groupsCould look at the two groups independentlyCaucasians Hispanics .263 .429.214 .319 .373 .487New Goals:Want to analyze the difference in the two proportions and get one confidence interval and one p-value measuring the strength of evidence that the difference in proportions is larger than we would expect by chance aloneNew descriptive statisticsNumerical and graphical summariesNew inferential methodsComparing two proportionsDefinitionWhen we have two variables in a study (e.g., ethnicity and whether overweight) we often consider one the explanatory variable and the other the response variable. Often the research study is looking for evidence that the explanatory variable causes changes in the response variable.(c) Can we draw a cause-and-effect conclusion here?Example: OK City ThunderTwo-way table Sell-outcrowdSmallercrowdTotalWin 3 12 15Loss 15 11 26Total 18 23 413/18 = .167 12/23 = .522Graphical summarySegmented bar graphConfounding variablessell-outGames comparesmaller crowdStronger opponentWeaker opponentKey IdeaObservational studies are always open to confounding variables and therefore we are not able to draw any cause-and-effect conclusions between our explanatory variable and our response variable with an observational study.ExampleLowerElevatingRandomizing Subjects appletBest Table in Entire CoursePotential for sampling bias Potential for confoundingSo what next?Can we eliminate “random chance” as an explanation?Pre-lab
View Full Document