IPHS402 - Analytical and Research Methods, Fall 2016 HOMEWORK 7 61 Points Due: 10/23/16 IPHS402 Analytic and Research Methods, Fall 2016 Page 1 Question 1: (20 points total) Doll and Hill, 1950: A study was conducted to evaluate the association between cigarette smoking and lung cancer. Cases of lung cancer consisted of all persons admitted to hospital with the diagnosis of lung cancer during a specified period. A total of 709 cases were identified. An equal number of controls were selected from other persons in the hospital at or about the same time who did not have a diagnosis of lung cancer. A \"smoker\" was defined as person who had smoked at least 1 cigarette a day for one year. Ninety-seven percent of cases were \"smokers,\" whereas 80% of controls were smokers. a. Draw the 2 x 2 table for these data (including labels for rows and columns) (4 points) b. Calculate the appropriate statistical measure of association and its 95% CI, and interpret your results. (4 points) c. Perform a chi-square test of association between smoking status and lung cancer. Show all steps and state your conclusions fully. Also, describe how your conclusions compare to those in the previous question. (6 points) d. You find out after your data are collected that the cases over-stated their smoking exposure and in fact only 90% were exposed. What is this form of bias called? (2 points) e. Recalculate the appropriate statistical measure of association given this new information (2 points) and discuss the impact on this association (2 points). Question 2 (12 points) In a small pilot study, 10 women with blood clots and 10 women without blood clots were contacted and asked whether they used hormonal contraception (HC). The women were matched by age, race and marital status. The results are shown below: Pair # Blood Clot Yes Blood Clot No 1 2 3 4 5 6 7 8 9 10 Used HC Used HC Did not Use HC Used HC Used HC Did not use HC Did not use HC Used HC Used HC Did not Use HC Did not use HC Did not use HC Used HC Used HC Did not use HC Did not use HC Did not use HC Did not use HC Used HC Used HC a. Draw the appropriate 2 X 2 table including labels for columns and rows (2 points). Calculate the estimated odds ratio of blood clots for using HC when analyzing this matched pairs study and interpret this odds ratio (3 points). b. Un-match the data above and redraw the 2 X 2 table with appropriate labels (2 points). Calculate the odds ratio associated with this unmatched data and compare this result to the matched odds ratio (is the difference in the direction you might have expected?) (3 points). c. This was a pair matched study. There is another often used method of matching in casecontrol studies. Name and describe this approach. (2 points) IPHS402 Analytic and Research Methods, Fall 2016 Page 2 Question 3 (28 points total) The data contained in (hw7_vietnam_data.txt) are derived from a national study of psychological symptomatology in Vietnam era veterans (I realize that many of you in this class were not born at the time of this war, but trust us, there was a Vietnam war - just a little humor here - although the war was certainly far from funny). The following two research questions are to be answered (using SAS) with these data (data dictionary below as well as the way to code the input line in your SAS code for data that are in columns): 1. Is service in Vietnam associated with an increased risk of psychological problems after discharge from active military duty? 2. Is level of combat exposure associated with an increased risk of psychological problems after discharge from active military duty? a) Generate age by using birth year subtracted from 1975, the year the Vietnam war ended and provide descriptive statistics for age. (2 points) b) State both the null and alternative hypotheses for each of the two research questions (4 points). c) Construct the 2X2 table for research question 1 above. Report and interpret the appropriate measure of association along with its 95% CI. (6 points) d) Evaluate the distribution of combat exposure (is it normally distributed?) using both a histogram and a Q-Q plot, which can be generated using PROC UNIVARIATE.(3 points) e) Based on your evaluation of the distribution of combat exposure create a dichotomous variable from combat exposure. Discuss your rationale for the decision you made (3 points). f) Construct the 2X2 table for research question 2 above. Report and interpret the appropriate measure of association along with its 95% CI. (6 points) g) You are concerned that branch of service may be a confounder of the association between combat exposure and risk of psychological problems after discharge. Explain how you would determine if branch of service fits the definition of a confounding variable? (2 points). If branch of service does fit the definition of a confounding variable, how would you address this in your analysis? (2 points) IPHS402 Analytic and Research Methods, Fall 2016 Page 3 Variable Names and Values for Vietnam Data Variable Column ID 1-6 Birth year 7-8 Year of release 9-11 Values or range 12 - 57 64-78 998 = unknown Drug/drinking problem 12 1 = yes 2 = no Branch of service 13 1 = army 2 = navy 3 = air force 4 = marine 5 = other Service in Southeast Asia 14 1 = yes 2 = no Combat Intensity Index 15-16 0 - 9 (least - most) 99 = no service in Southeast Asia data hw; infile 'c:\\LOCATION of data file\ ame of data file; input CASEID 1-6 BRYEAR 7-8 RELYEAR 9-11 DRUGDRNK 12 BRNCHSV 13 SESERV 14 COMBAT 15-16; if relyear = 998 then delete;***this is done to eliminate those observations with unknown year of release. The rest of your program is placed here. IPHS402 Analytic and Research Methods, Fall 2016 Page 4 a. Perform a chi-square test of association between smoking status and lung cancer. Show all steps and state your conclusions fully. Also, describe how your conclusions compare to those in the previous question. (6 points) Lung cancer No lung cancer Total Smokers 560 (688) 567 1127 (1,255) Non smokers 149 (21) 142 291 (163) Total 709 709 1418 H0: there is no association between smoking and cancer Ha: There is no association between smoking and cancer We calculate the expected frequencies A11= 709*1127/1418= 564 (627.5) A12=709*1127/1428=564 (627.5) A21=709*291/1418=146 (81.5) A22=709*291/1418=146 (81.5) E O (E-O)2/E 564 (627.5) 560 0.4539 564 (627.5) 567 0.01596 146 (81.5) 149 0.0616 146 (81.5) 142 0.1096 =0.64106 Degrees of freedom= (2-1)(2-1) =1 Level of significant is 0.05 From the table critical chi square, with 1 degree of freedom and 0.05 level of significant is given by: X21 ,0.05=3.841 Calculated chi-square is less than the critical chi-square. That is 0.64106<3.841 We therefore do not reject the null hypothesis. Conclusion We conclude there is no sufficient evidence to conclude that there is association between smoking and cancer. a. Perform a chi-square test of association between smoking status and lung cancer. Show all steps and state your conclusions fully. Also, describe how your conclusions compare to those in the previous question. (6 points) Lung cancer No lung cancer Total Smokers 560 (688) 567 1127 (1,255) Non smokers 149 (21) 142 291 (163) Total 709 709 1418 H0: there is no association between smoking and cancer Ha: There is no association between smoking and cancer We calculate the expected frequencies A11= 709*1127/1418= 564 (627.5) A12=709*1127/1428=564 (627.5) A21=709*291/1418=146 (81.5) A22=709*291/1418=146 (81.5) E O (E-O)2/E 564 (627.5) 560 0.4539 564 (627.5) 567 0.01596 146 (81.5) 149 0.0616 146 (81.5) 142 0.1096 =0.64106 Degrees of freedom= (2-1)(2-1) =1 Level of significant is 0.05 From the table critical chi square, with 1 degree of freedom and 0.05 level of significant is given by: X21 ,0.05=3.841 Calculated chi-square is less than the critical chi-square. That is 0.64106<3.841 We therefore do not reject the null hypothesis. Conclusion We conclude there is no sufficient evidence to conclude that there is association between smoking and cancer