Mathematics 243 (18.16) We are assuming that a packet behaves like a random sample of all seeds (Randomization Condition) and that the package is less than 10% of the whole population (10% Condition). This latter is true since the population is all possible seeds produced by this company. If we use the normal approximation, we must also check the Success/Failure condition (8% of 160 is 13). We can compute this probability exactly using the binomial or approximately using the normal approximation. > 1-pnorm(.95,.92,sqrt(.92*.08/160)) [1] 0.0809429 > 1-pbinom(.95*160,160,.92) [1] 0.05275305 (18.18) Computation using both the binomial and the normal approximation to the binomial. March 19 Homework Solutions that there is a 90% probability that this particular interval contains the parameter. Our confidence is in the procedure, not the particular interval. 90% of all 90% confidence intervals will contain the true value of the parameter. (19.24) (a) Randomization Condition: random sample is stated. 10% condition: There are many more that 17,000 college freshmen. Success failure condition: 74% of 1644 and 26% of 1644 are both greater than 10. (b) > phat=.74 > se = sqrt( .74*.26/1644) > zcrit = qnorm(.99) > c( phat- zcrit*se, phat+ zcrit*se) [1] 0.7148333 0.7651667 (c) We are 98% confident that the national freshmen-tosophomore retention rate is between 71.5% and 76.5%. (d) In 98% of random samples, the 98% confidence interval captures the true proportion so we can be 98$ confident that this interval does. (19.28) (19.22) (a) The parameter being estimated is the percentage of legal songs in the digital music libraries of all students. > phat=49/207 The population is songs and the sample size is 117,079. > se=sqrt ( phat * (1-phat) /207) (b) Certainly the 10% Condition and Success failure con> zcrit = qnorm(.95) dition are satisfied. The songs sampled are not a simple > c( phat- zcrit*se, phat+ zcrit*se) random sample but rather a cluster sample of all songs. [1] 0.1881192 0.2853107 This means that independence is rather badly violated. (a) A 90% confidence interval for the success rate of this (c) clinic is (.188, .285). > phat=.231 (b) We can be 90% confident that the true success rate > se = sqrt ( phat * (1-phat) /117079) of the clinic is in this interval. > zcrit=qnorm(.975) (c) 90% of all confidence intervals constructed by this > c( phat- zcrit*se, phat+ zcrit*se) method capture the true value of p. [1] 0.2285858 0.2334142 (d) It is consistent with this interval that the true success rate is 25% as 25% is in the interval. (d) We can be 95% confident that between 22.9% and 23.3% of all student songs are legal. While this confiIt is very easy to incorrectly describe what a confidence dence interval was correctly constructed, it is quite small interval means. Read page 472 to get a sense of what due to the large sample (and subsequent small standard a confidence interval does not mean! It is important to error). We wouldn’t necessarily trust it to this level of realize that a probability statement is always about the confidence especially because of the cluster sample rather future. So don’t say that a 90% confidence interval means than the SRS. > 1-pbinom(19,732,.04) [1] 0.973135 > 1-pnorm(19/732,.04, sqrt(.04*.96/732)) [1] 0.9737477