Chapter 1 : Review Exercises (Errors may have crept in during the scanning process.)

1.94. Letter-writers feel strongly about the issue, most often strongly negative. They may be very unrepresentative of the opinion of all constituents. We can't tell from the letters what the majority opinion is.

1.95. Voluntary response is the big reason. Opponents of gun control usually feel more strongly than supporters, and so are more likely to call. The sampling method also reduces response from poorer people by requiring a phone and willingness to pay for the call.

1.96. Readers of this magazine are certainly atypical to begin with. Within this group we have voluntary response again. The claimed benefits are subjective ("some benefit"), so it is easy for enthusiasts to respond positively. All in all, the 93% is certainly too high.

1.97. (a) American college students. (b) The PSY 001 class list. (c) It is not representative of the population, since only freshmen psychology students at one school are listed. (d) It is a bit slanted in favor of commercials. The question suggests that commercials are needed if we are to have TV. Asking if the subject prefers commercials to other means of paying for TV would get a lower percentage favoring commercials. (e) Aside from (c) and (d), "are in favor of TV commercials" does not accurately reflect the question actually asked, and no margin of error is given even though the sample is very small. (f) Not at all.

1.98. (a) "Adult women," probably female U.S. residents age 18 and over. (b) Statistic. (c) We are 95% confident that between 70% and 82% of adult women think an "elegant" car is highly desirable. (This is based on the 6% margin of error from Table 1-1, the most conservative choice.)

1.99. The margin of error would be (very nearly) the same. Margin of error is determined by sample size; population size has an effect only if the population is not very large relative to the sample.

1.100. (a) E.g., In 19 of 20 such samples the results would lie within three percent age points of the percent of all adults who jog." (b) The margin of error includes only random sampling error, the variation due to chance in selecting a sample at random. Not included are errors due to, e.g., inability to contact selected individuals (nonresponse) or the respondent lying (response error).

1.101. (a) Residents of Alaska and Hawaii are omitted, as are all those in households without phones. (b) We are 95% confident that between 63% and 69% of 'call" adults favor a school prayer amendment. (In class, ask students to say what 95% confidence means.) (c) Practical difficulties include all sources of nonsampling errors: inability to contact the selected telephone numbers, subjects refuse to respond or misunderstand the question. If the target population is all adults, omitting the groups in (a) is a source of sampling error as well.

1.102. Answers may vary (especially in how they are written for someone "who knows no statistics"), but should be based on the definitions: (a) A simple random sample (of size n) is n units (people, tennis balls, TVs, etc.) chosen in such a way that every collection of n units in the sampling frame has the same chance of being selected. (b) A statement made with 95% confidence is one that is based on the outcome of a procedure that works (gives correct results) 95% of the time. (c) Nonsampling error are errors in a survey that arise from something other than the act of choosing a sample; for example, processing errors, nonresponse, response errors, and poorly worded questions. (d) Human subjects in a study should give informed consent before collecting data. This means that they must be told what effect the study might have on them (how much time it will take, whether there is any risk, etc.), and then must agree to accept those conditions.