Suggested problem set #4 Chapter 7: 4, 9, 10, 11, 17, 20 Chapter 7. 4. For the following scenarios, say whether the binomial distribution would describe the probability distribution of possible outcomes. If not, say why not. a. The number of red cards out of the first five cards drawn from the top of a regular deck. No, because the probability of getting a red card would change as the cards are removed. The binomial assumes that the probability of success is equal and independent for each trial. b. The number of red balls out of ten drawn one by one from a vat of 50 red and blue balls, if the balls are replaced after each draw. Yes c. The number of red balls out of ten drawn one by one from a vat of 50 red and blue balls, if the balls are not replaced after each draw. No, for the same reason as in question a. d. The number of red flowers in a square meter plot in a field that has been randomly strewn with an indefinitely large amount of seed. No, because the number of trials is not specified. e. The number of red eyed flies among 50 Drosophila individuals drawn at random from a large population. yes f. The number of red eyed flies in five Drosophila families, each of ten individuals, with the families chosen at random from a large population. 9. a. p ˆ = 1856 5743 =0.323 no, because the individuals are not chosen at random. b. No, because this is not a sample it describes the entire population. Any error in this estimate (and there is likely some) is due to measurement error, not sampling variation. 10. Ho: The proportion of species moving north is p=0.5. Ha. The proportion of species moving north is not p=0.5. Use a binomial test. P = 2 * (Pr[x = 22] + Pr[x = 23] + Pr[x = 24]) P = 0.000002 We reject the null hypothesis, and conclude that most species are moving north.
11. a. The first, because the confidence interval is smaller. b. The first, for the same reason c. No, because their confidence intervals overlap, and the confidence interval for the second study includes the mean from the first study. 17. Carry out a binomial hypothesis test. Ho: The probability of choosing an OM female is p=0.5. Ha: The probability of choosing an OM female is different from 0.5. Test statistic: 19 out of 24 choose OM P= 2*(Pr[x =19] + Pr[x = 20] + Pr[x = 21] + Pr[x = 22] + Pr[x = 23] + Pr[x = 24]) P = 0.0015 Reject the null hypothesis. There is evidence that male choice is affected by fetal positioning. b. Worse, because then the two females in each trial would not be independent. 20. a. 30% b. Binomial c. The same as the standard deviation of a binomial distribution: n * p(1" p) = 15*0.3(0.7) =1.77 n=15 here because it s the number of cells picked per researcher, and not the number of researchers, that determines the s.d. d. p(1" p) = 0.3(0.7) = 0.118 n 15 e. 95% Chapter 8: 3, 6, 8, 14 3. a. Six categories 1, 2, 3, 4, 5, 6. Five degrees of freedom (binomial dist.) b. Data are the number of heads in 10 tries. 11 categories 0, 1, 2, 3, 4, 5, 6, 7, 8, 9, or 10 heads. 10 degrees of freedom (binomial dist.). c. Same number of categories. Here you d have to estimate the parameter p, so there s one less degree of freedom 9. d. 5 Categories: 0, 1, 2, 3, 4. 3 degrees of freedom (5-1-1). 6. a. Ho: Birth rates on weekdays and weekends are equal. Ha: Birth rates differ between weekdays and weekends. b. Binomial: difficult to calculate but exact, assumes random sample Chi-squared goodness of fit test easier to calculate, requires large sample size to be accurate. c. Ho: Birth rates on weekdays and weekends are equal.
Ha: Birth rates differ between weekdays and weekends. Use chi-squared goodness of fit test, with the proportional model. Observed Expected Weekdays 716 665.7 Weekends 216 266.3 Total 932 932 Expected for weekdays: 5/7 * 932 = 665.7; weekends: 2/7 * 932 = 266.3 Test statistic: (O # E) 2 (716 # " 2 665.7)2 = $ = + E 665.7 df = 2-1 = 1 Critical value at alpha = 0.05: 3.84 (216 # 266.3)2 266.3 =3.80+9.50=13.3 The test statistic is larger than the critical value, so we reject the null hypothesis. 8. Use a chi-squared goodness of fit test for a Poisson distribution. Ho: Number of parasites follow a Poisson distribution Ha: Number of parasites follow some other distribution. Calculate the mean: (0*103+1*72+2*44+3*14+4*3+5*1+6*1)/238=225/238=0.945 Use this in the Poisson formula to calculate the expected values: Number of parasites Frequency Expected 0 103 92.51 1 72 87.42 2 44 41.30 3 14 13.01 4 3 3.07 5 1 0.58 6 1 0.09
This violates the assumptions of chi-squared, so we lump the last three categories: Number of parasites Frequency Expected 0 103 92.51 1 72 87.42 2 44 41.30 3 14 13.01 >4 5 3.75 Now calculate the chi-squared statistic: 4.58 df = 5 1 1 = 3 Critical value with alpha = 0.05, 3 df: 7.81 Fail to reject the null hypothesis there is no evidence that fish differ in their ability to attract nematodes. 14. " a. Pr[x = 8] = 15 % $ '(0.5) 8 (0.5) 7 # 8 & b. Use a goodness-of-fit test. Wins Observed Binomial Expected 0 0 0.00 0.07 1 7 0.00 0.98 2 25 0.00 6.83 3 56 0.01 29.60 4 112 0.04 88.81 5 190 0.09 195.39 6 290 0.15 325.64 7 256 0.20 418.68 8 549 0.20 418.68 9 339 0.15 325.64 10 132 0.09 195.39 11 73 0.04 88.81 12 56 0.01 29.60 13 22 0.00 6.83 14 22 0.00 0.98 15 3 0.00 0.07 2132 We have to lump the first two and last two categories. Wins Observed Expected Chi-squared 0-1 7 1.04 34.11 2 25 6.83 48.32 3 56 29.60 23.54 4 112 88.81 6.05 5 190 195.39 0.15
6 290 325.64 3.90 7 256 418.68 63.21 8 549 418.68 40.56 9 339 325.64 0.55 10 132 195.39 20.56 11 73 88.81 2.82 12 56 29.60 23.54 13 22 6.83 33.68 14-15 25 1.04 552.00 2132 852.98 The critical value for a chi-squared with 13 df is 22.36. We reject the null; these data do not fit a binomial distribution. c. You can see from the above plot that the observed number of wrestlers with exactly 8 wins is much higher than expected. d. There is an alternative that some wrestlers are just better than others. This would mean that for each match, p is not equal to 0.5. The plot provides some support for this in that there are way more contestants with 0-1 wins (very bad) and 14-15 wins (very good) than expected. Chapter 9: 5, 8 5. (84+72)/5865 pairs were of different species.
so: p ˆ = 156 5895 = 0.026 SE p ˆ = p ˆ (1" p ˆ ) n "1 = 0.026(1" 0.026) 5895 "1 = 0.0021 Do a chi-squared contingency test. Observed: Female Collared Pied Row Sums Collared 5567 84 5651 Pied 72 172 244 Column sums 5639 256 5895 Expected: Female Collared Pied Row Sums Collared 5405.6 245.4 5651 Pied 233.4 10.6 244 Column sums 5639 256 5895 Test statistic: (O " E) # 2 = 5895. E Chi-squared distribution with (r-1)(c-1) = 1 d.f. Critical value: 3.84. Our test statistic is greater than the critical value, so we reject the null hypothesis. The birds are pairing nonrandomly. 8. a. Expected Number of matings 1 2 3 4 5 Gave birth 82.37 88.05 57.75 16.10 4.73 249 Didn't give birth 4.63 4.95 3.25 0.90 0.27 14 87 93 61 17 5 263 b. No there are cells with expected values less than one, and more than 20% with expected less than 5. Categories need to be pooled to meet the assumptions. c. Observed Number of matings 1-2 3-5
Gave birth 166 83 249 Didn't give birth 14 0 14 180 83 263 Expected Number of matings 1-2 3-5 Gave birth 170.42 78.58 249 Didn't give birth 9.58 4.42 14 180 83 263 Test statistic: # (O " E) 2 E = 6.82 Chi-squared distribution with (r-1)(c-1) = 1 d.f. Critical value: 3.84. Our test statistic is greater than the critical value, so we reject the null hypothesis. d. No both could be correlated with a third variable, like female condition. Females who are in better shape plenty of food, not sick might get more mates and also have larger litters.