In: Statistics and Probability
3. (Not Bayesian) Let Y have pdf fY (y, θ) =
2(θ − y)/θ2
if 0 < y < θ
0 otherwise . You are
going to construct a confidence interval for θ based on a single
observation Y .
(a) Show that Y/θ is pivotal quantity.
(b) Suppose Y = 6. Find (numerically) a 10% confidence interval for
θ .
The given pdf of Y is
Let
. When y=0, u=0 and when
Then
and
and
the pdf of U is given by
So the pdf of
does not depend on
and hence
is a pivotal quantity.
Let a be the point such that
Then
implies
Therefore
will be a 10% confidence interval for
.
Now we will find the value of a.
We know P(a<U<1)=0.1
That is
or
Solving this quadratic equation we get the value of a as
So
and the 10% confidence interval becomes