\documentclass[12pt]{article}
\usepackage{curves}
\begin{document}
\large
\begin{center}
Midterm Exam
Psychology 406
Fall 2007
\end{center}
Name: \rule{5in}{.02in}
\bigskip
Whenever possible, use fractions throughout.
\newpage
I. ELISA tests are used to screen donated blood for the presence of the AIDS (HIV) virus. The test actually detects antibodies, substances that the body produces when the virus is present. When antibodies are present, ELISA is positive with probability 0.997 and negative with probability 0.003. When the blood tested is not contaminated with AIDS antibodies, ELISA gives a positive result with probability 0.015 and a negative result with probability 0.985. Suppose that 1\% of a large population carries the AIDS antibody in their blood.
The following results may be useful in answering the questions below:
\[ P(A) = P(A|B)P(B) + P(A|\bar{B})P(\bar{B}) \]
\[ P(A|B) = \frac{P(B|A)P(A)}{P(B|A)P(A) + P(B|\bar{A})P(\bar{A})} \]
(1) What is the probability that the ELISA test for AIDS is positive for a randomly chosen person from this population?
\vspace{2.0in}
\newpage
(2) What is the probability that a person has the antibody given that the ELISA test is positive (for a randomly chosen person from this population)?
\vspace{2.0in}
(3) What is the probability that a person doesn't have the antibody given that the ELISA test is positive (for a randomly chosen person from this population)?
\vspace{2.0in}
(4) What is the joint probability that a person doesn't have the antibody and the ELISA test is positive (for a randomly chosen person from this population)?
\vspace{2.0in}
\newpage
II. Suppose I have a set of observations, $X_{1},\ldots,X_{N}$, and
I turn them into $Z$-scores, $Z_{1},\ldots,Z_{N}$. If I carry out a
further transformation to $T_{i} = 10 Z_{i} + 50$, what is the
resulting mean and variance of $T_{1},\ldots,T_{N}$?
\vspace{2.0in}
Besides just giving the answers, demonstrate how you got them using our rules for summations (and how I might have done this in class).
\vspace{2.0in}
\vspace{2.0in}
\vspace{2.0in}
\newpage
III. Suppose $X$ is a random variable that takes on the following values with the probabilities listed:
\bigskip
P($X$ = -1) = 10/32; P($X$ = 0) = 20/32; P($X$ = 1) = 1/32; P($X$ = 2) = 1/32
\bigskip
a) E($X$):
\bigskip
b) Var($X$):
\bigskip
c) Mode:
\bigskip
d) Median:
\bigskip
e) E($X^{2}$):
\bigskip
f) Standard deviation of $X$:
\bigskip
g) P($X \ge 0$):
\bigskip
h) E(-3$X$ - 7):
\bigskip
i) Var(-3$X$ - 7):
\bigskip
j) $X$ represents a``fair'' game (yes or no):
\newpage
IV. Let $X$ be a continuous random variable with the cumulative
distribution function given above.
\bigskip
\begin{figure}
\begin{center}
\setlength{\unitlength}{1.0in}
\begin{picture}(2,3)
\put(0,0){\line(0,1){2.0}}
\put(0,0){\line(1,0){3.0}}
\put(.5,-.1){\line(0,1){.2}}
\put(1.0,-.1){\line(0,1){.2}}
\put(1.5,-.1){\line(0,1){.2}}
\put(2.0,-.1){\line(0,1){.2}}
\put(2.5,-.1){\line(0,1){.2}}
\put(.5,-.2){\makebox(0,0)[t]{1}}
\put(1.0,-.2){\makebox(0,0)[t]{2}}
\put(1.5,-.2){\makebox(0,0)[t]{3}}
\put(2.0,-.2){\makebox(0,0)[t]{4}}
\put(2.5,-.2){\makebox(0,0)[t]{5}}
\put(-.1,.2){\line(1,0){.2}}
\put(-.1,1.6){\line(1,0){.2}}
\put(-.1,.4){\line(1,0){.2}}
\put(-.1,.6){\line(1,0){.2}}
\put(-.1,.8){\line(1,0){.2}}
\put(-.1,1.0){\line(1,0){.2}}
\put(-.1,1.2){\line(1,0){.2}}
\put(-.1,1.4){\line(1,0){.2}}
\put(-.2,.2){\makebox(0,0)[r]{1/8}}
\put(-.2,.4){\makebox(0,0)[r]{2/8}}
\put(-.2,.6){\makebox(0,0)[r]{3/8}}
\put(-.2,.8){\makebox(0,0)[r]{4/8}}
\put(-.2,1.0){\makebox(0,0)[r]{5/8}}
\put(-.2,1.2){\makebox(0,0)[r]{6/8}}
\put(-.2,1.4){\makebox(0,0)[r]{7/8}}
\put(-.2,1.6){\makebox(0,0)[r]{8/8}}
\put(0,0){\line(5,2){1.0}}
\put(1.0,.4){\line(5,6){.5}}
\put(1.5,1.0){\line(5,2){.5}}
\put(2.0,1.2){\line(5,4){.5}}
\put(2.5,1.6){\line(1,0){.5}}
\end{picture}
\end{center}
\end{figure}
\newpage
a) Calibrate the pointer above to correspond to the cumulative
distribution function.
\bigskip
\begin{figure}
\begin{center}
\setlength{\unitlength}{1.0in}
\begin{picture}(2,2)
\put(1,1){\bigcircle{2}}
\end{picture}
\end{center}
\end{figure}
\bigskip
b) P($X$ = 2):
\bigskip
c) P($3 \le X \le 5)$:
\newpage
V. As part of an article that appeared in the \emph{NY Times}
(10/9/2003) (which is attached), and after Arnie won the California
recall, a section is given on ``How the Poll was Conducted''
beginning on the second page. Explain and show what the section
means, in terms of formulas you know, about the margin of error in
estimating population proportions. Does the plus or minus two points
require any rounding up to two decimal places?
\vspace{6.0in}
\newpage
VI. On a particular standardized reading test, the mean score for
50 low-achieving students was 61.3 with a standard deviation
of 10.2. According to the publisher's manual, the mean score
in the general population should be 70. For now, assume that
the population from which the low-achieving students were drawn
can be represented by a random variable $X$ that is N($\mu$,$\sigma^2$).
If necessary, use interpolation in the tables to obtain the
appropriate $t$-value for use (and show how you did this).
\bigskip
(a) To assess whether the population mean for the low-achieving
group might differ from that of the general population
used to standardize the reading test, carry out a test of
$H_{o}: \mu = 70$ versus $H_{1}: \mu < 70$ at a fixed alpha level of .01.
\vspace{2.0in}
(b) Construct a 99\% confidence interval for the unknown mean
$\mu$ for the low-achieving population.
\vspace{2.0in}
\newpage
(c) Assuming that the standard deviation in the low-achieving population is a known value of 10.2, what sample size would be required to have a 99\% confidence interval be 4 points in total length? Show your work.
\newpage
VII. In a taste test of two versions of cola (e.g., pepsi versus
coke), I asked 6 people independently which they prefer
(blindfolded and randomly deciding on each trial whether pepsi or
coke is given first to the person to taste). Suppose the
preference for pepsi is represented by a random variable $Y$ taking
on a value of 1 and the preference for coke represented by the
random variable $Y$ taking on a value of 0.
\bigskip
We let $\hat{p}$ denote the observed proportion of the 6 people who prefer pepsi to
coke, and assume in the population from which these people are
drawn that P($Y$ =1) = 1/2.
\bigskip
(a) Find the probability distribution for $\hat{p}$:
\bigskip
Values of $\hat{p}$ \hspace{10ex} Probability
\vspace{3.0in}
(b) E($\hat{p}$) =
\bigskip
(c) Var($\hat{p}$) =
\bigskip
\newpage
VIII. I have sampled 10 hyperactive children and have blood pressure
scores before and after the administration of a tranquilizer.
The data are as follows:
\bigskip
\begin{tabular}{ccc}
Subject & Before & After \\
1 & 155 & 110 \\
2 & 160 & 120 \\
3 & 140 & 145 \\
4 & 145 & 125 \\
5 & 155 & 145 \\
6 & 160 & 115 \\
7 & 150 & 120 \\
8 & 175 & 155 \\
9 & 170 & 145 \\
10 & 165 & 140 \\
\end{tabular}
\bigskip
Suppose $\mu_{B}$ and $\mu_{A}$ refer to the population means for the before and
after scores, respectively.
(a) Using the usual t-test procedure for dependent samples, carry
out a test of $H_{o}: \mu_{B} - \mu_{A} = 0$ versus $H_{1}: \mu_{B} - \mu_{A} > 0$ at a fixed
alpha level of .01.
\vspace{3.0in}
(b) Construct a 95\% confidence interval on $\mu_{B} - \mu_{A}$.
\vspace{2.0in}
(c) Using the sign test, state the hypothesis you would now test
that would be analogous to what as carried out in (a).
Provide the exact $p$-value.
\vspace{2.0in}
\newpage
IX. I am interested in the effects of a certain chemical that is
typically sprayed on fruit to preserve it during transit, and
decide to carry out an experiment on maze learning behavior in
rats to see if the chemical might have any noticeable influence.
Given the 36 rats I have available, 18 are randomly assigned to a
``chemical additive'' diet and 18 to a regular diet. Some of the data follow, where the dependent variable is the number of trials
it takes a rat to make five consecutive successful runs
(no rat was given more than 20 trials). Use the summary
statistics to answer the two questions, (a) and (b).
\bigskip
\begin{tabular}{cccc}
Chemical Additive Diet & & Regular Diet & \\
7 & 20 & 8 & 9 \\
8 & 9 & 9 & 13 \\
12 & 14 & 10 & 20 \\
20 & 16 & 14 & 18 \\
$\vdots$ & $\vdots$ & $\vdots$ & $
\vdots$ \\[2ex]
Sum: & 270 & Sum: & 222 \\
Sum of Squares: & 4430 & Sum of Squares: & 2950 \\
\end{tabular}
\bigskip
The assumptions are made that the data in the (independent) groups I and II came from a $N(\mu_{1},\sigma_{1}^{2})$ and $N(\mu_{2},\sigma_{2}^{2})$, respectively.
\newpage
(a) Give the 95\% confidence interval on $\mu_{1} - \mu_{2}$ using the
approximate degrees-of-freedom calculated from the data. If necessary, use interpolation in the tables to obtain the appropriate $t$-value for use (and show how you did this).
\vspace{3in}
(b) Carry out a test of $H_{o}: \mu_{1} - \mu_{2} = 0$ versus $H_{a}: \mu_{1} - \mu_{2} \ne 0$, using the
``pooled'' version of the $t$-test at a fixed alpha level of .05. Again, if necessary find the appropriate $t$-value for use by
interpolation (and show how you did this).
\vspace{3in}
\newpage
X. Completion
\bigskip
a) The sampling distribution for the number of
successes in $n$ independent observations on a
population characterized by a dichotomous outcome
of success/failure, is called the \rule{2.0in}{.02in}.
\bigskip
b) If I have 6 individuals and need to form a
committee of 3 from this group, how many different
ways could this be done? \rule{2.0in}{.02in}
\bigskip
c) The sample proportion $\hat{p}$, since E($\hat{p}$) = $p$, is said
to be a(n) \rule{2.0in}{.02in} estimator of $p$.
\bigskip
d) An estimator $\hat{\theta}$ is said to be \rule{2.0in}{.02in} if $\hat{\theta}$ converges to $\theta$ asympototically.
\bigskip
e) The robustness of several hypothesis testing
procedures is based on the fact that sample means
are approximately normal when simple random
samples are taken and the sample size gets large. This result is called the \rule{2.0in}{.02in}.
\bigskip
f) Generally, and other things being equal, power
decreases both as the sample size \rule{2.0in}{.02in}, and as
the alternative hypothesis value being considered
gets \rule{2.0in}{.02in} the value specified under $H_{o}$.
\bigskip
g) One-tailed hypothesis tests are also referred to
as \rule{2.0in}{.02in} hypothesis tests.
\bigskip
h) In calculating power for a specific value
within the alternative hypothesis, 1.0 minus the power is called the \rule{2.0in}{.02in}.
\bigskip
i) The estimated degrees-of-freedom for a two-independent samples $t$-test with 10 observations in
one group and 15 in the second must fall between
the values of \rule{2.0in}{.02in}and \rule{2.0in}{.02in}.
\bigskip
j) The assumption in the ``pooled'' two-independent
sample $t$-test that $\sigma_{1}^{2} = \sigma_{2}^{2}$ appears to not be very
important (i.e., we have robustness) as long as
\rule{2.0in}{.02in}.
\bigskip
k) 5 ``factorial'' is equal to \rule{2.0in}{.02in}.
\bigskip
l) If $X_{1} \sim N(0,5)$ and $Y_{1} \sim N(1,2)$ and $X_{1}$ and $Y_{1}$
are independent, then $3X_{1} - 4Y_{1}$ is \rule{2.0in}{.02in}.
\bigskip
m) 95\% confidence intervals are ``naturally connected''
with fixed $\alpha$-level two-tailed tests, where $\alpha$ is
equal to \rule{2.0in}{.02in}.
\bigskip
n) If the \rule{2.0in}{.02in} is as small or smaller than $\alpha$,
we then typically say that the data are statistically significant
at level $\alpha$.
\bigskip
o) The standard error of a statistic is also called
the \rule{2.0in}{.02in} of the statistic.
\bigskip
p) The variance of a $t$-distribution with $k$ degrees-
of-freedom converges to \rule{2.0in}{.02in} as $k$ goes to
infinity.
\bigskip
q) A continuity correction is sometimes used when the
probabilities from a binomial distribution are
approximated by probabilities from a(n)
\rule{2.0in}{.02in}.
\bigskip
\newpage
XI. I have a test statistic $T$ that has the following distribution
under $H_{0}$ and $H_{1}$:
\bigskip
\begin{tabular}{ccc}
$T$ & Under $H_{0}$ & Under $H_{1}$ \\
1 & 1/8 & 2/8 \\
2 & 2/8 & 2/8 \\
3 & 2/8 & 2/8 \\
4 & 2/8 & 1/8 \\
5 & 1/8 & 1/8 \\
\end{tabular}
\bigskip
If you have a decision rule that says to reject $H_{0}$ if $T$ = 1,
2, or 3, find:
\bigskip
a) Type I error:
\vspace{1in}
b) Type II error:
\vspace{1in}
c) Power:
\vspace{1in}
\end{document}