Stat 217 – Day 27Last Time – Inference for RegressionExample 2: Textbook pricesExample 1: Gesell dataSlide 5Slide 6In factSlide 8Influential ObservationsExample 2: FEVSlide 11Stat 217 – Day 27Topics in RegressionLast Time – Inference for RegressionHo: no association or =0Ha: is an/positive/negative associationMinitab outputt test statistic from “coefficient of x” rowtwo-sided p-valueSE(b) = amount of random variation of slopes from sample to sampleStrength of evidence (p-value) vs. strength of association (r)P. 605Example 2: Textbook pricesThe relationship between textbook price and number of pages appears stronger than between textbook prices and year of pub.Every time the pages are increased by one the predicted price will increase by $.147.The p-value is 0.000 (t = 7.65) so I would reject the null hypothesis. Strong evidence there is an association between pages in textbook and price (in the population).Example 1: Gesell dataCan we predict later intelligence based on when the child first speaks?Removing the one child has a pretty big impact on the regression line, significance Example 1: Gesell dataExample 1: Gesell dataBest conclusion?Some evidence that children who take a particularly long time to speak may have lower IQ scores, but otherwise no relationship between age of first words and later IQ.For children who take between 5 and 20 months, no relationship…In fact42 months and Gesell = 120Regression line follows that one child!Example 1: Gesell data15 months and Gesell = 57Influential ObservationsObservations whose removal has a dramatic impact on the regression line (or p-value or correlation coefficient)Most likely candidates – extreme x valuesExample 2: FEVTo Turn in with partnerFor WednesdayLab 9Start bringing review questions(Thursday in lab – case study type
View Full Document