Chi Square Fr Chris - St Francis High School March 2006
Two Applications Goodness of Fit (one way) Two-Way (Contingency) Table
1-Way (Univariate) Data Observed frequencies are from a RANDOM SAMPLE Large Sample size (expect at least 5 in each category)
Stolen Cars by Color Null Hypothesis: Color makes no difference San Luis Obispo Telegram-Tribune 9/2/96 Alternative Hypothesis: Color makes a difference 15% white, 15% blue, 35% red, 30% black, 5% other whiteblueredblackother Observed Expected
whiteblueredblackother Observed Expected SRS, and All EXPECTED cells >5 so large enough sample Since P(Chi-sq, df=4 > 18.46) <.001 so we reject the null hypothesis, and conclude certain colors are more likely to be stolen
Now you try: According to SLO Trib, 12/15/99, SRS of 200 purchases of California Lotto tix: Age played California in 1999 had 35% 18-34, 51% 35-64, 14% over 65 Try a Chi-Sq Goodness of Fit Test
2 way Chi-Square is even more Fun! According to Research Quarterly for Exercise and Sport, (1990) p , SRS of 1200 looked at hours of TV and a cardiovascular test Hrs of TVFitUnfit
But we compute df=(r-1)(c-1) and the expected frequencies are computed with by the marginal totals Hrs of TV FitUnfitTotals
Hrs of TV FitUnfitTotals Expected Frequency = (Row)(Col)/total df=(4-1)(2-1)=3, So we reject the Null Hypothesis at the.05 level for any Chi Square over 12.83
less than 12.83, so there is not enough evidence to reject the null hypothesis at the.05 level, so there is little evidence here that would suggest that the hours watching TV is associated with one’s cardiovascular fittness.
sexdiedlived men women classdiedlived high middle low Would you rather be rich or a woman if you were on the Titanic? Actually the Chi-Square statistic tests against the null hypothesis (no difference according to category)... you need to look at the numbers to see what’s going on...