You are on page 1of 2

What You Should Know for the Midterm

About models that we have covered: You will not be expected to memorize
equations that we have derived in class, such as the equation of the ROC curve
[in the case of signal detection theory]. You should know what these relationships
are about: for example, that the ROC curve describes the relationship between one
variable and another [the hit rate and the false alarm rate]; that its numerical
specification depends on a particular parameter of the underlying theoretical model
[the value of d]; what kind of relationship it shows [upward sloping, above the
diagonal, concave]; how it is affected by changes in model parameters [moves
farther up and to the left of the diagonal, the higher is d].
About mathematical results or definitions that we have used in class: You should
know the meaning of basic statistical concepts [mean, median and variance of a
random variable; conditional probability, conditional mean, conditional variance;
posterior probability distribution and Bayes Rule; bias of an estimator, mean
squared error of an estimator; likelihood ratio test and sequential probability ratio
test as approaches to deciding between two hypotheses] and be able to apply them
in simple examples. You will not be expected to memorize more complex definitions
such as the equation for the normal density function, though you should know what
the probability density function of a random variable tells you about it.
An important result about normally distributed random variables that you should
know [as we have now used it repeatedly] is the answer to the following question:
suppose that x is a normally distributed random variable, and y = x + e, where e is
another normally distributed random variable [with mean zero], independent of x;
then what is the conditional expectation of x, conditional on the value of y? [This
comes up when y is a noisy observation of some true state x, or y is the imprecise
subjective perception caused by some external stimulus x, and we want to know
what estimate of x should be formed on the basis of an observation of y.] You also
should know basic results from probability theory about the means and variances
of sums of independent random variables. Any numerical or algebraic calculations
required on the exam will, however, be fairly simple. You are allowed to bring a
calculator to the exam, but it should not strictly be necessary, since in the case of
any calculations, the point will be to indicate how you should do the calculation,
not what the actual numerical result is.
About references that we have discussed: You will not be expected to memorize the
names of the authors of the individual studies, let alone the dates or other details of
that kind; nor any precise numerical details of the authors findings. You should
know what the main point was of each of the studies that we have discussed in
some detail [though not necessarily of studies that were mentioned only very briefly
in passing]. For example, you should know that in the study of undergraduates
ability to predict how happy they would be if assigned to different residential houses
[it was by Dunn, Wilson and Gilbert, but you dont need to remember that], subjects

assigned to houses that they had ranked above average before the assignment
later reported themselves to be less happy with their residential house than they
had expected to be. You would not be expected to remember that there were 12
different houses, that the hypothesis that they would be as happy ex post as they
had predicted ex ante was rejected at the 1 percent level, or small details of that
kind.

You might also like