1 252thngs 5/3/01 ECO 252 - THINGS THAT YOU SHOULD NEVER DO ON A STATISTICS EXAM (OR ANYWHERE ELSE)! x and squaring it. 1. Compute x 2. Compute xy by multiplying x by y . 2 by computing 3. Claim that a point below the 50th percentile is above the median, or that a point above the 50th percentile is below the median. 4. Claim that a variance ( 2 or s 2 ) or a standard deviation is negative (or that the variance or standard deviation of x is zero unless x is a constant). 5. Claim that a sum of squares is negative. a. Claim that x 2 nx 2 is negative. b. Think that because x 2 nx 2 can't be negative, neither can xy nxy . c. Take the square root of -1.(unless you are taking a very advanced course where i 1 is legal.) 6. Claim that a probability (or a p-value) is above 1 or negative. 7. Claim that a large p-value means that a coefficient is significant or that a small p-value means that it is insignificant. 8. Compute a sample statistic or a population parameter without checking the value of n or N . 9. Claim that a correlation (r ) is above 1 or below -1 or that a coefficient of determination ( R 2 ) is above 1 or negative. 10. Claim that population means, variances, medians or proportions differ (i.e. that a difference is significant) without doing a hypothesis test. 11. Claim that parameters (e.g. population means) differ because sample statistics (e.g. sample means) differ. 12. Do a hypothesis test without stating H 0 and H 1 . 13. State a null hypothesis that does not include an equality, or an alternative hypothesis that includes an equality. 14. State a null or alternative hypothesis that contains a sample statistic such as a sample mean, sample variance or sample proportion. These belong in your tests, not your hypotheses. 15. Use a two-sided confidence interval or critical value to test a one-sided hypothesis. 2 252thngs 5/3/01 16. When asked a question about the median, answer a question about the mean (unless you are sure that the distribution is symmetrical). 17. When a distribution is not normal, use a sample median to test a hypothesis about a median. (This sort of hypothesis is always tested by using a method using ranks or proportions.) 18. When asked for coefficients (b1 , b2 ) in a multiple regression, use the formula for b1 in a simple regression. In other words, deciding that, since b1 b2 Sx1 y in simple regression, it must be true that SSx1 Sx 2 y in multiple regression won't get you an ounce of credit for this type of problem. SSx2 19. When using a non-parametric method that requires ranked data, fail to rank the data or even to check whether it is ranked. 20. When using data that are paired or cross-classified, fail to realize that cross classification usually implies a different method from that used for independent random samples. 21. When using data from independent random samples, assume that, because lines are numbered or because two samples are of the same size, that the data is cross-classified. 22. Believing that, because the data in a problem is similar to a problem that you have seen before, that the answer is similar. Read the problem first! 23. When the alternate hypothesis is 0 : a) using a critical value that is above 0 ; b) using a positive value of t ; or c) using a confidence interval or when the alternate hypothesis is 0 : a) using a critical value that is below 0 ; b) using a negative value of t ; or c) using a confidence interval. 24. When asked if the mean is above or greater than 0 , assuming that the null hypothesis is 0 or 0 (Instead the alternate hypothesis is 0 !) or when asked if the mean is below or less than 0 , assuming that then null hypothesis is 0 or 0 (Instead the alternate hypothesis is 0 !) -------------------------------------------------------------------------------------------------------------------------------Of course, due to rounding errors or arithmetic mistakes, some of the above things will occur. If they do, look for your error . If you can't find it, don't panic. Your answer may still be good for whole or partial credit if you include a note on your paper that you realize that your result is impossible.