HSA 523 Homework #4 Answer Key Dr. Robert Jantzen Economics Department 1. Given the contingency table, and let the two risk factors of low-birth weight and illegal drugs be denoted as LB & ID: Had Illegal Drugs: Yes No Total Had Low-Birth Weights: Yes No 30 20 50 200 80 220 Total 50 250 300 What’s the probability that an infant will be born w/ low-birth weight? P(LB) = 80/300 = .267 What’s the probability that an infant will be born w/ illegal drugs in their blood? P(ID) = 50/300 = .167 B. What’s the joint probability that an infant will have both low birth weight and blood w/ drugs in it? P(LB&ID) = 30/300 = .1. This is the chance of being born w/ both risk factors. What’s the joint probability that an infant will have neither? P(LB’ & ID’) = 200/300 = .667. This is the chance of being born w/ neither risk factor. C. What’s the conditional probability that a low birth weight infant will have drugs in their blood? P(ID|LB) = 30/80 = .375. This is the chance a low birth weight infant will be born w/ illegal drugs. What’s the conditional probability that an infant w/ drugs in their blood will have low birth weight? P(LB|ID) = 30/50 = .6. This is the chance an infant w/ drugs in their blood will be low birth weight. A. 2. Given P(HIV) = .02, P(HIV & Male) = .015, and P(Male) = P(Female) = .5: A. What’s the joint probability of being infected w/ HIV and being a women? (HINT: simple probability is = to the sum of the joints) Since P(HIV) = P(HIV&Male) + P(HIV&Female), then .02 = .015 + P(HIV&Female), so the joint probability of being infected w/ HIV and being a women infant, i.e., P(HIV&Female) is .005. B. What’s the conditional probability of being infected w/ HIV for men? (HINT: conditional is = to joint / simple) P(HIV|Male) = P(HIV&Male) / P(Male) = .015 / .5 = .03. Men have a 3% chance of being HIV infected. What’s the conditional probability of being infected w/ HIV for women? P(HIV|Female) = P(HIV&Female) / P(Female = .005 / .5 = .01. Women have a 1% chance of being HIV infected. What’s the conditional probability of an HIV infected person being a man? P(Male|HIV) = P(HIV&Male) / P(HIV) = .015 / .02 = .75. The chance that you’re a male if you’re HIV+ is 75%. What’s the conditional probability of an HIV infected person being a woman? P(Female|HIV) = P(HIV&Female) / P(HIV) = .005 / .02 = .25. The chance that you’re a female if you’re HIV+ is 25%. D. HIV infection and gender are not statistically independent because the simple probability of infection is not equal to the conditionals, i.e., P(HIV) is not equal to P(HIV|Male) or P(HIV|Female). Men are more likely (.03) to be infected than women (.01). C. 3. Given that 10% of transplant patients die after surgery, 30% survive w/ marginal improvement, and 60% completely recover, and alcohol abusers comprise 30% of deaths, 20% of those w/ marginal improvements, and 10% of complete recoveries. We can form a Bayes Table to answer the questions because the first three percentages are the simple probabilities after surgery and the latter three are conditional probabilities of alcohol abuse for each of the surgical outcomes. Let P(SOi) indicate the surgical outcome probabilities, and P(AB|SOi) the probabilities of alcohol abuse for each surgical outcome. Surgical Outcome Death Marginal Improvement Complete Recovery Prior prob. of each Conditional prob. of alcohol abuse for surgical outcome P(SOi) 0.1 each surgical outcome P(AB|SOi) 0.3 0.3 0.6 0.2 0.1 Joint prob. of alcohol abuse for each surgical outcome P(AB&SOi) 0.03 Revised conditional prob. for each surgical outcome given 0.06 0.06 0.15 .4 .4 alcohol abuse P(SOi|AB) .2 Given the above: A. What fraction of all transplant patients abuse alcohol? 0.15 which is the sum of all of the joint probabilities of abusing alcohol and the surgical outcomes. B. Given a patient abuses alcohol, what are the probabilities that they’ll die, show marginal improvement or make complete recoveries? The probabilities are .2, .4 & .4. There’s a 20% chance of death for an alcohol abuser, a 40% chance of marginal improvement and only a 40% chance of complete recovery. 4. Given that a hospital knows that 5% of their inpatients will suffer a “hospital caused” complication and 95% won’t, and weekend admissions comprise 3% of those w/ complications and 2% of those w/o complications. We can form a Bayes Table to answer the questions because the first two percentages are the simple probabilities of complications & no complications, and the latter two are the conditional probabilities of being admitted on the weekend for each of these initial outcomes. Let P(COMPL?i) indicate the two initial outcome (complication & no complication) probabilities, and P(WKEND|COMPL?i) the probabilities of weekend admission for each outcome. inpatient outcome Conditional prob. of weekend admission for each inpatient outcome P(COMPL?i) 0.05 0.95 P(WKEND|COMPL?i) 0.03 0.02 Prior prob. of each Inpatient Outcome Complication NoComplication Joint prob. of weekend admit for each Revised conditional prob. for each inpatient outcome given inpatient outcome weekend admit P(WKEND&COMPL?i) 0.0015 0.019 0.0205 P(COMPL?i|WKEND) .073 .927 A. What’s the simple probability of suffering a complication? .05 Not suffering a complication? .95 B. What’s the simple probability of being admitted on the weekend, i.e., what fraction of all inpatients were admitted on the weekend? 0.0205 which is the sum of all of the joint probabilities of being admitted on the weekend and the two inpatient outcomes, i.e., had complications or didn’t.. C. Given an inpatient was admitted on the weekend, what’s the probability that they’ll suffer a complication? Won’t suffer a complication? The probabilities are .073 & .927. There’s a 7.3% chance of complication for a weekend admission (& 92.7% chance of no complication), which is 50% higher than the general 5% complication rate.