General Instructions
To do each example, just click the Submit
button.
You do not have to type in any R instructions or specify a dataset.
That's already done for you.
Theory
The theory of the parametric bootstrap is quite similar to that of the nonparametric bootstrap, the only difference is that instead of simulating bootstrap samples that are IID from the empirical distribution (the nonparametric estimate of the distribution of the data) we simulate bootstrap samples that are IID from the estimated parametric model.
All the same considerations arise.
- Since the parameter estimate
theta hat
is not the true parameter valuetheta
. We do not sample from the correct distribution. We should sample from Fθ. We do sample with the same thing with ahat
on the θ (which I can't do on a web page). - Thus the bootstrap does not do the right thing, only close to the right
thing when the sample size is large.
- In constructing confidence intervals, it helps to bootstrap pivotal or at least variance-stabilized quantities.
And so forth.
Simulating from a parametric model is not so easy as simulating from the
empirical distribution. In fact, it can be arbitrarily complicated. So
hard that it is an open research problem
how to do it. For some
parametric models sampling is easy, others not. In general, it bears no
relation to sampling from the empirical.
If the observed data are in the vector x
, then
x.star <- sample(x, replace = TRUE)
makes a nonparametric bootstrap sample.
In contrast, if the observed data are assumed to be IID normal, then
x.star <- rnorm(length(x), mean = mean(x), sd = sd(x))
makes a parametric bootstrap sample. This does not do the right thing
because we should specify mean
to be the true population mean
and sd
to be the true population standard deviation (but since
we don't know the population values we must use estimates).
For more contrast, if the observed data are assumed to be IID Cauchy, then
x.star <- rcauchy(length(x), location = median(x), scale = IQR(x) / 2)
makes a parametric bootstrap sample. We can't use mean(x)
and sd(x)
as estimators of location and scale because the
Cauchy distribution doesn't have moments and hence these aren't consistent
estimators (of anything, much less location and scale).
Why median(x)
and IQR(x) / 2
are consistent
(even asymptotically normal) estimators of location and scale would be
more theory than we want to go into here. The only point we wanted to
make is that the three examples look a lot different from each other.
The Normal Location Problem
Comments
- We are using a trimmed mean
mean(x, trim = 0.25)
for the location estimator (on-line help). The untrimmed mean is optimal for the normal distribution (assuming we really believe the data are normal), but we can use whatever estimator we want and the bootstrap is still valid. - We are using the median absolute deviation
mad(x)
for the scale estimator (on-line help). The standard deviation is optimal for the normal, but . . . .
The Cauchy Location Problem
- We are using a trimmed mean
mean(x, trim = 0.25)
for the location estimator (on-line help). No simple estimator is optimal. - We are using the median absolute deviation
mad(x)
for the scale estimator (on-line help). No simple estimator is optimal. We change theconstant
argument to be right for the Cauchy distribution. The population MAD being given byqcauchy(0.75)
, which is 1.0.
Theory II: Multinomial Sampling
To get to some examples with wading through a tremendous amount of theory, we will stick to one parametric model for which the sampling looks fairly similar to the nonparametric bootstrap. This is the multinomial distribution.
The multinomial distribution is the distribution of categorical measurements
on IID individuals. The number of individuals in each category make up
the data vector x
and the probabilities of individuals being
in each category make up a probability vector p
(where probability vector
means all(p >= 0)
and
sum(p) == 1
).
Given a probability vector p
of length k
and a sample size n
one creates a multinomial sample with the R statements
c.star <- sample(1:k, n, prob = p, replace = TRUE) x.star <- tabulate(c.star, k)
(The first statement creates an IID sample of category numbers with
the specified probabilities. The second counts the number of individuals
in each category. So x.star
is a vector of length k
.)
A Chi-Square Test of Goodness of Fit
For example, suppose we observe the multinomial data defined to
be x
in the form below, and we want to test the null hypothesis
that the true category probabilities are all equal (to 1 / 6 because there
are 6 categories). The R function chisq.test
does the usual
chi-square test that uses the large-sample approximation (that the chi-square
test statistic has a chi-square distribution). The remainder of the code
does the parametric bootstrap test.
Actually, since the null hypothesis
is completely specified here this is, strictly speaking, a Monte Carlo
test rather than a parametric bootstrap. The test is exact.
(We only say we are doing a parametric bootstrap
when the
null distribution of the test statistic depends on the true unknown
parameter value and we are plugging in an estimate for the unknown truth).
Goodness of Fit with Estimated Parameters
For another example, suppose we observe the Poisson data defined to
be x
in the form below, and we want to test the null hypothesis
that the data is Poisson against the alternative hypothesis that is isn't.
The usual way to do this is to truncate the data, making a category
8 or more
for example. The reason for the truncation is that
the usual asymptotics of the chi-square test requires it. But this
introduces needless complication when we are not using the asymptotics
(except that theta hat
is near theta
).
Comments
- The reason why
tabulate(1 + x, 1 + max(x))
is thatx
contains integers starting at zero buttabulate
wants integers starting at one. - The test statistic calculated by the
tstat
function is the likelihood ratio test statistic. By standard likelihood theory, it is asymptotically equivalent to the chi-square test statistic, but is not exactly the same for any finite sample size. - The reason why
foo[counts == 0] <- 0
is that whencounts[i]
is zero, then so isp1[i]
, but that makesfoo[i]
zero times minus infinity equalsNaN
, which is not right. Since x log(x) converges to zero as x goes to zero, we wantfoo[i]
to be zero in that case. - Interpretation of the P-value: P = 0.34 means
no evidence against H0. Hence we accept the
null hypothesis that the data are Poisson.
(A
goodness of fit
test like this is just like every other hypothesis test except that we want to accept H0 as opposed to the usual situation where we want to reject. Our acceptance of H0 doesn't mean that the data actually have the Poisson distribution, only that they provide no reason to disbelieve they do.)
A Chi-Square Test of Independence
For another example, suppose we observe the contingency table defined to
be x
in the form below, and we want to test the null hypothesis
of independence (that the row category labels and column category labels
are independent random variables).
For some reason, the R chisq.test
function does this test
for 2-way tables by simulation, but doesn't do the analogous test for
1-way tables (go figure).
Logistic Regression
Theory
The glm
function in R
(on-line
help) does so-called generalized linear models
.
The theory of these is usually covered in the categorical data course
(like Stat 5421), so we won't cover it here.
For those interested, I wrote up some notes about generalized linear models for another class (Stat 5931), but there's no real need to look at it to understand this example.
The response variable in this problem kyphosis
is categorical with values present
or absent
which we model as independent but not identically distributed
Bernoulli random variables
where the so-called mean value parameter vector
is related to the so-called linear predictor vector
by the link function
and the linear predictor vector has the usual form of a mean function in ordinary linear regression
where X is the so-called design matrix or model matrix for the problem (the rows are cases and the columns are predictor variables) and β is a vector of regression coefficients.
Kyphosis is a misalignment of the spine. The data are on 83 laminectomy
(a surgical procedure involving the spine) patients. The predictor variables
are age
and age^2
(that is, a quadratic function
of age), number
of vertebrae involved in the surgery
and start
the vertebra number of the first vertebra involved.
The response is presence or absence of kyphosis after the surgery
(and perhaps caused by it).
Practice
Comments
- The command
pred <- predict(out2, type = "response")
estimates the mean value parameter vector p. Iftype = "link"
(the default) it produces the linear predictor η, which is not what we need for the next item. - We use predicted values for
out2
because that is the fit for the small model (null hypothesis) and you always simulate under the null hypothesis when doing a test. - The command
kyphosis.star <- rbinom(n, 1, pred)
simulates new Bernoulli data with mean value parameter vector p. That's what the parametric bootstrap requires. -
We save both the bootstrap values of the test statistic
dev.star
and the P-valuepev.star
(which can also be thought of as a test statistic with the proviso that we reject the null for low values ofpev
as opposed to high values ofdev
and other usual test statistics). -
The plots serve as a diagnostic of whether the parametric bootstrap is
necessary and also show what it does.
-
In the first plot
hist(dev.star)
the histogram should approximate a chi-square distribution with 2 degrees of freedom, but this is a bit hard to see. -
In the second plot
hist(pev.star)
the histogram should approximate a uniform distribution on the interval (0, 1), and it is clear that, although the approximation is not too bad, it could be better. -
In the third plot, the theoretical Q-Q plot of
pev.star
against the quantiles of the uniform(0, 1) distribution, again the approximation is not bad but could be better. In this last plot, the operation of the parametric bootstrap is to use the curve rather than the straight line in computing the (bootstrap, corrected) P-value. Hence the difference between the curve (which is the empirical CDF of the bootstrap P-values) and the straight line (which is the CDF of the uniform(0,1) distribution) at the abscissa corresponding to the ordinary P-value derived from believing the asymptotics.
-
In the first plot