This preview has intentionally blurred sections. Sign up to view the full version.
View Full DocumentThis preview has intentionally blurred sections. Sign up to view the full version.
View Full DocumentThis preview has intentionally blurred sections. Sign up to view the full version.
View Full DocumentThis preview has intentionally blurred sections. Sign up to view the full version.
View Full Document
Unformatted text preview: PubH 6450 NAME: E " 5
Exam 1 (10/08/09) Lecture Instructor (please circle): Susan Telke Lynn Eberly Directions: This is an open—book, opennotes exam; however, sharing of books, notes, handouts, homework or lab
papers, calculators, or verbal comments is not permitted. You may use a calculator of your choosing, but
laptop computers and any kind of internet connection are not permitted. For multiple choice questions, please clearly indicate your answer(s). For short answer problems, please show all relevant work necessary to arrive at a solution, as partial
credit will be awarded. DO NOT LOOK AT OR BEGIN THE EXAM UNTIL
THE INSTRUCTOR ANNOUNCES THAT THE EXAM
IS STARTING. You will have from 1:25pm to 3:20pm to complete this exam, so do not spend too much time on any
one problem. With such a big class, there is always someone who needs to take the exam early or late. After the
exam is over, DO NOT DISCUSS THIS EXAM with anyone who was not in this same room taking it
with you. We anticipate returning the exams in class on Tuesday. Good luck! ! Problem 1: Information on cause—ofdeath is lacking for 98% of the world's 4 million neonatal deaths that
occur in countries with inadequate vital registration (VR). The aim of this study was to estimate,
for the year 2000, the distribution of neonatal deaths across multiple causes including: asphyxia,
preterm birth, congenital abnormalities, infection (sepsis/pneumonia), neonatal tetanus, diarrhea,
and ‘other’. A group of 44 countries with adequate VR was used to estimate death rates for 56
countries with inadequate VR. Estimates were obtained by combining death information with
several other variables, including female literacy rate, categorized as under 40%, 4080%, and
above 80%. [Source: Lawn, et al. International Journal of Epidemiology 2006; 35(3): 706—718] Use this information to answer questions A thru G. A. What is the study design?
a. Clinical Trial
b. Case—Control c. Cohort Study ' pr» 2, Crosssectional Study Ema1mm, D 0U;
Watt: N07 FOLLau/ZD N0 MH'I'CI—IING ofé Eyl/Poswrez/ Np gﬁNO0MIEA‘r/0nl. cor/N1IZIEIS
R 'T/mE'l USED! B. For each variable below indicate if it is quantitative, nominal, ordinal or binary. 77‘ a. Infant Birth Weight (grams) Cat/A N7 IT” 1/ u: b. Reason for Neonatal Death: asphyxia, preterm birth, congenital abnormalities,
infection, neonatal tetanus, diarrhea, and ‘other’ MOM/IV ﬁl d. Outcome (death or survival) BIA/"7'22 {0"’”"’“ 0Q "OM/"4‘
ak “7m ) ‘1
2, 0. Female Literacy Rate Quﬂﬂ‘rl‘fﬂﬂ It
Q, [0 Below are box plots, separately for each cause, showing the distribution across countries of
proportion of deaths using (a) Adequate VR data (44 countries, some causes missing) and (b)
Inadequate VR data (56 countries, all causes of interest shown). a 4 (a) .s t ...... H G! Proportion 01 deaths IN ADE o un—r :
UK,
C. For Plot 21 (Adequate VR data), is the proportion of neonatal deaths due to congenital
abnormalities Left Skewed, Right Skewed, or (circle one)? D. For Plot b (Inadequate VR data), which measure of central tendency is most appropriate
for the proportion of neonatal deaths due to congenital abnormalities? 2/ Minimum, Mean, r Variance (circle one) [$727me Y3 E. In the context of this study, what can you say about the median and variability of
neonatal death rates due to preterm birth compared to those "due to asphyxia across both
the adequate VR data and the inadequate VR'data groups” [C "' 7 Z/g’ p’czdj/QCQ’ZT/O/‘j NEG ,«Vﬁngﬂ pgﬂT/rs Our: 7‘0 H§17HVWA Mars Lam/£72
THg/v DUE 70 PIZETE'Zm 8/2774 For: é$oTH
ﬂOEQU/g‘fé‘ MAJ/o 1wnmEQu/9'rg‘ We ‘7‘/—/E‘ 9,97” G/ZOU/Ds. a“774,5” t/ﬂrZ/ﬂiglLivy 0F NED/V9704. 0Fr9‘7'H5
19W; 70 PiZETE/Zﬂ? Bria—7H was (exaggng
7714/“ V/q’z’i/Q’BILITV 0F i’VFO/VI‘97WL_,
pgrQT/VLS pug 7c x45/7lIV'X/I—7 Pom 807”
Qmoups ‘ i _ I “ V l B .01'HEIZ. UﬂI/LI/OT/O/ZS /4 CCE P774ng F. The proportion of neonatal deaths due to preterm delivery in the United States follow a
normal distribution with a mean value of p: 3.6 deaths per 1000 deliveries and a standard
deviation of o :05 deaths per 1000 deliveries. Use the SAS output from the program
below to answer the questions a through e. data norm; u = cdf ('normal', 3.8, 0 1);
v = cdf ('normal', ~1.2, 0, 1);
w = cdf ('normal', 5.6, 0.F, 3.6),
x = Cdf ('nonmal , 3.0, 0.5, 3.6),
y = probit (0.&0);
z = probit (0.90);
run; ~———+——4———+"“*"_T_—l__—l_——‘
proc print data = norm; .3) 3'“ "No C\
2! 74!
run;
Obs u v w x y E .
/ Z
1 0.99744 0.11507 0.89435 0.75630 1.28155
1.28155 a. What is the probability that randomly selected community from this population has a
neonatal death rate due to preterm deliveries between 3 deaths per 1000 deliveries and 5
deaths per 1000 deliveries? = 14,4 :— .aqawq— 415°? ¢ 0 . 8 3 2. 3 7
b. What neonatal death rate due to preterm deliveries is the cutoff for the lowest 10% of the
distribution?
2 r {2%
6"
X — 3 (a ....7,9SS"‘ (»\.28tSSX05) +39 =1 x lb ' x4295? c. Suppose we repeatedly take samples of size 300. What is the sampling distribution of the sample mean death rate due to preterm deliveries, Y ? (Give the name of the distribution
and the values of its mean and standard deviation.) n: 300 Y N N(A,U/VZ) ) where, Mr—Z(l
Wm: G's/G50 Y~N(;.c)o.om) "—0.019 d. A randomly selected community from an inadequate VR data group showed a sample mean neonatal death rate due to preterm delivery of X = 5 deaths per 1000 deliveries.
Construct a 95% confidence interval for the true mean death rate in this population,
assuming a standard deviation of 0 :05 deaths per 1000 deliveries. Note that from SAS probit(0.975) = 1.96. lb G. Below is a table of the number of neonatal deaths due to different causes of neonatal
mortality for a random sample of women from India (n=618). a. Create a table that shows the probability distribution for neonatal death from asphyxia
(yes/no) for a randomly selected participant in this study. 3? 15,18 :— 0.1408
CiroHot?) = 08577. b. Define S as the sample space. For a randomly selected participant in this study, is it
true that:
P(S) = P(A) + P(B) + P(C) + P(D) +P(E) + P(F) + P(G) = 1? Why or why not?
\lEs; nu, gag/v15 HRE 0/5JO’NT “"40
Evgk‘f Poss/5 Lt E VENT A5
Con/5 IDEQED .
0. Suppose the true proportion of deaths due to asphyxia is 0.12 for the population represented by this sample. What is the sampling distribution for, f) , the proportion of neonatal deaths in this study from asphyxia? (Give the name of the distribution and the values of its mean and
standard deviation.) 9: 0.;2. np =(913Yotz) = 74.“, >
harp): 57435“! 507” "'0
l\ P(pp . :o.\‘z_
g P N N(P) 1—)>) where 1; 8)
"'"" ' 1‘3
%NNL°'1110 08) n
6 PMNN 7, Problem 2: 866 patients who recently experienced a heart attack are recruited for a study. Each patient was
categorized into one of four New York Heart Association (NYHA) classes, which indicate the
amount of activity that person is able to do without angina; the patients were then followed for
mortality. The researchers are interested in time to death and how it might vary by NYHA class. Use this information to answer questions A thru D. Circle the correct answer for each of A through D: rospective Randomized A. This study is: Retrospective I" servationa ' NYHA class B. This study is: Time to death @ C. The exposure of interest is: D. The outcome of interest is: Problem 3: A prospective cohort study of patients with primary biliary cirrhosis of the liver was done at the
Mayo Clinic. Researchers were interested in time to death after onset of treatment. Patients and
their physicians could choose from among three treatment options: prednisone, ursodeoxycholic
acid (UDCA), or both. Researchers also recorded age at each patient’s first diagnosis of primary
biliary cirrhosis. The objective of the study was to compare time to death across the three
treatment groups. Use this information to answer questions A and B. A. Age at diagnosis is related to time to death: those who have been living with the disease
longer are likely to have more advanced disease. Age at diagnosis may also be related to
treatment: those who have been living with the disease longer are more likely to choose the dual
therapy (both prednisone and UDCA). What is the epidemiological term for this kind of variable
when examining the association of treatment with death? 0/2 Lf/ﬂklﬂé VHWH$ L? B. The researchers are therefore worried that the distribution of the age at diagnosis values
might impact their conclusions. How could this prospective study been done differently to avoid this concern? , ﬁgznxzwaes Cal/AD lZ—ﬂ/VﬁDMIZE pﬁV/E/W5 IN7O TREﬂTmE’V‘f ézatxps I» f/QPEFl/Lt—y I MaﬁrvarI/G 77/5
ﬁ,g pjsm/BI/r/on/ Arc/2055 émups_ @ . Msgﬂwgzzs ace/Lo 5' way 77/; EFFECTS 0F "774E
ﬂEﬁ'T/ﬁ EW75 WNW/M ﬂéE r éRovps [/pw EVEK/
B 5ﬁm/LE 5/2; Mﬂf 85 W /551/£ . 7 Problem 4: A recent study examined the use of ultrasound (US) to correctly diagnose appendicitis in 283
children. The ultrasound result indicates whether or not appendicitis was detected.
Histopathology was used to determine the true appendicitis status. The data were as follows: D+ D"
__ True disease status 
—_ Diagnostic 94 103
T+ test result a endicitis
US indicates no 165 180
T“ a endicitis
Let D+ denote the true presence of appendicitis. Let T+ denote a US test that indicates
appendicitis.
Use this information to answer questions A thru E.
A. What is Pr(D+)? o)
O = .
Fr (94 3 c: ‘ /2%3 0 395'
3
B. What is Pr(T+)?
03 
P, (Ti) a ' /283 ~ 0'5“?
3
C. What is the sensitivity of the US test for detecting appendicitis?
PCT+ and D0 cwI183 Qq
T+D“')= ——————“"—'——————= r: .8692.
P ( mm ) m, A03 0
283
E
D. What is the specificity of the US test for detecting appendicitis?
T « and D’) NS 23 b
17(12 my): P—Q/ __— I 3 = Li» : 0.948
E PLD') mtg/2.83 W
E. What is the positive predictive value of the US test for detecting appendicitis? Use the back of
this a e if ou need more 5 ace.
pg y pPCﬁoMU> q‘i/283 :94A03goag
PLTt 3
L‘ on
PD? \Dﬂ PU») ’ ﬂ!“
mm \rﬂ — a“ W,” PU,» +€CTHV3PLD5 .... ., B Mew“3 8 Q9) (0.8623(0335 )‘l’ (o‘oszyo'mg) (03¢ 13/ Emmowe
2229K) Problem 5: In worker residences next to a large mine, the probability of finding mercury in the household
water supply is 0.55. In this same neighborhood, the probability of having diminished responses
to a neurological exam is 0.40, and the probability of either finding mercury or having
diminished responses is 0.85. EV g N—r A ; mgrzcu 27‘ QUENT g '. DMMNtsHeD
RESPON 96$ A. Find the probability that a person in this neighborhood has mercury in their household water
supply and has a diminished neurological response. Use this information to answer questions A thru C. PCA and 8): FLA) r PLBU’PLAOQBQ == 040 B. Find the probability that a person in this neighborhood does not have mercury in their
household water supply and does not have a diminished neurological response. 02 M5? demw DIAbQIQM C. In this neighborhood, is having mercury in the household water supply independent of having
a diminished neurological response? Justify your answer. No') WA M e\ 4% PLM PUB) PUMBB it PM) .40
PCI’AM eff”) =rL ss’ ...
View
Full Document
 Fall '10
 AndyMugglin

Click to edit the document details