Statistics 401E Fall 2010 SECOND EXAM Solutions Name _______________________________ The Gay Liberation Movement (GLM) is generally thought of as having begun on June 17, 1969, at the Stonewall Inn—a gay bar in New York City. Although police raids of such bars were commonplace at the time, this date marks the first instance when police officers met violent resistance from homosexual men while attempting to raid a gay bar. Along with the Women’s Movement, GLM has sought to undermine gender-related discrimination in the US. One strategy endorsed by both movements has been toward the reduction of stereotypical depictions of male and female demeanor in the mass media. In Gender Advertisements (1976) Erving Goffman provided numerous illustrations of such gender-stereotypical depictions within advertisement photography. For example, in magazine advertisements women’s hands are typically depicted as “barely touching, even caressing” objects, whereas men’s hands are more typically “grasping or manipulating” them. Men instruct women, but not the reverse. Women may sit or lie on a bed, but not men. Many studies have subsequently examined US magazine advertisements, and have found in them pervasive evidence of such gender-cues. However, all of the magazines studied until now have been “straight” ones (i.e., magazines with advertisements that target traditional heterosexual consumers). You decide to fill this gap in the research by studying advertisements in lesbiangay-bisexual-transgender (LGBT) as well as straight magazines. From among all issues published between 1970 and 2010 of the straight magazine, Glamour, and the LGBT magazine, Advocate, you randomly sample 121 issues from each magazine. You then select the first instance of advertisement photography in each of these issues, yielding a sample of 242 photographs—121 from the straight magazine, Glamour, and 121 from the LGBT magazine, Advocate. Your first hypothesis is that advertisement photography in straight magazines (e.g., Glamour) is more likely to have gender-cues than photography in LGBT magazines (e.g., Advocate). Your selection of the 1970 starting date is due to its proximity to the beginning of GLM. Since you believe that GLM (along with the Women’s Movement) has been effective, your second hypothesis is that the number of gender-cues in the two magazines has declined linearly over the past four decades. You have data on the following three variables: YEAR (Y) = Year of the issue in which the photograph appeared NUMCUE (N) = Number of gender-cues depicted in the photograph TYPEMAG (T) = 1 if photograph is from the straight magazine, Glamour 0 if photograph is from the LGBT magazine, Advocate 1 Statistics 401E Fall 2010 SECOND EXAM Solutions Name _______________________________ Correlations, means, and standard deviations among these three variables are as follows: ——Correlation Coefficients—— Standard YEAR NUMCUE TYPEMAG Mean Deviation YEAR 1.00 –.05 .00 1990.0 11.8 NUMCUE –.05 1.00 .24 6.0 1.5 TYPEMAG .00 .24 1.00 0.5 0.5 (Notes: Please be sure that you understand these data. For example, the correlation between YEAR and NUMCUE is –.05 for your sample of 242 photographs. The mean value of YEAR is 1990 years; the standard deviation of YEAR is 11.8 years.) More data: On average there are 6.24 gender-cues (with a variance of 1.52) among the 121 photographs from Glamour. On average there are 5.55 gender-cues (with a variance of 2.49) among the 121 photographs from Advocate. a. Using the .05 significance level, test the hypothesis that the two variances listed in the just-mentioned bulleted sentences (namely, 1.52 and 2.49) are significantly different. Show your work and state your conclusion in words! [weight 4] 120 120 F ˆ A2 2.49 120 2 1.64 1.35 F120 ,.05 ˆG 1.52 Thus the variances are significantly different. b. What null and alternative hypotheses would you use to test your first hypothesis as stated in the third paragraph of this problem? (Hint: Remember to take into consideration the levels of measurement of the variables you will be using from the data provided above.) [weight 2] H o : N G N L H A : N G N L Here N G is the population mean value on the NUMCUE variable among photos from Glamour, and N L is the population mean NUMCUE value among photos from Advocate. 2 STAT 401E / SECOND EXAM c. Using the .05 significance level throughout, test the null hypothesis stated in part b. (Hint: Be sure to show your work and to state your conclusion in clear English!) [weight 5] Given the finding in part a separate variance estimate should be used. 2 ˆ SEPARATE And thus 1.52 2.49 4.01 .033 121 121 121 ˆ SEPARATE .033 .182 . Regarding to the hypotheses in part b, z ˆ SEPARATE 1.645 .182 .299 . The rejection rule is thus, reject Ho if X N G X N L .299 . Since X NG X N L 6.24 5.55 .690 .299 , we reject H o and conclude that during 1970-2010 gender-cues in first instances of advertisement photography were significantly more likely in Glamour than in Advocate. d. After regressing NUMCUE on YEAR, you obtain the following slope: ˆ 1 .5 bˆY rNY N .05 .0064 State in plain English (i.e., as discussed in ˆY 11.8 class) what this slope means. [weight 3] One would estimate for each subsequent year between 1970 and 2010 a decrease of .0064 in the number of gender-cues in Glamour and Advocate magazines’ first instances of advertisement photography. 3 STAT 401E / SECOND EXAM e. What null and alternative hypotheses would you use to test your second hypothesis as stated in the third paragraph of this problem? [weight 2] Hypotheses about a slope or a correlation may be used here. H o : bY 0 H A : bY 0 , where b̂Y is from Nˆ aˆ bˆY Y OR H o : NY 0 H A : NY 0 f. What rejection rule would you use in testing the hypotheses stated in part e of this problem at the .05 significance level? (Hint: Be sure to show your work, and to remember that a rejection rule is a sentence that always begins with the phrase, “Reject the null hypothesis if…”.) [weight 5] If part e states hypotheses about a slope: SS N n 1 ˆ N2 241 1.52 542.25 SS TOTAL SSY n 1 ˆY2 241 11.82 33,556.84 2 SSTOTAL 1 .05 542.25 540.894 SS ERROR 1 rNY ˆ 2 MSE ˆ b2ˆ Y 2 SS ERROR 540.894 2.254 n k 1 240 2.254 MSE .000067 thus ˆ bˆ Y 33,556.84 SSY .000067 .0082 z.05ˆ bˆ 1.645 .0082 .0135 Y Reject Ho if bˆY .0135 . OR If part e states hypotheses about a correlation: ˆ 2 1 rNY 1 .05 .0042 thus ˆ rNY .0042 .0645 n2 240 2 2 rNY z.05ˆ rNY 1.645 .0645 .1061 Reject Ho if rNY .1061 . 4 STAT 401E / SECOND EXAM g. Apply the rejection rule obtained in part f, and using clear English explain your conclusion in terms of the second hypothesis. [weight 1] I do not have statistically significant evidence (at α=.05) of a linear decline in gender-cues in the 2 magazines’ photographs over the past 4 decades, because bˆY .0064 .0135 (or, equivalently, because rNY .05 .1061 ). h. If between 1970 and 2010 there were actually a decline of .01 gender-cue per year in the two magazines, what is the probability that you committed a Type 2 [weight 3] error in part g? As always, show your work! z .0135 .01 .425 .0082 Pr(z<-.425)=Pr(z>.425)=.3354 β=Pr(Type II error)=1-.3354=.6646 Hypothetical truth 6646 -.01 -.0135 Rejection region Acceptance region 5