Professional Documents
Culture Documents
Estimation
Helge Blaker & Emil Spjtvoll
Raiden Hasegawa
University of Pennsylvania
A Road Map
A Road Map
MOTIVATING PROBLEM:
A Road Map
MOTIVATING PROBLEM:
Interval estimation procedures can yield paradoxical results, even in
simple settings.
A Road Map
MOTIVATING PROBLEM:
Interval estimation procedures can yield paradoxical results, even in
simple settings.
PLAN:
A Road Map
MOTIVATING PROBLEM:
Interval estimation procedures can yield paradoxical results, even in
simple settings.
PLAN:
1. Conceptual limitations of standard inferential methods.
A Road Map
MOTIVATING PROBLEM:
Interval estimation procedures can yield paradoxical results, even in
simple settings.
PLAN:
1. Conceptual limitations of standard inferential methods.
2. A potential graphical alternative.
A Road Map
MOTIVATING PROBLEM:
Interval estimation procedures can yield paradoxical results, even in
simple settings.
PLAN:
1. Conceptual limitations of standard inferential methods.
2. A potential graphical alternative.
3. An illustrative example.
The Setup
The Setup
Suppose you have data X from some model P and youd like to
conduct two-sided inference on R.
The Setup
Suppose you have data X from some model P and youd like to
conduct two-sided inference on R.
The standard workhorses for inference:
The Setup
Suppose you have data X from some model P and youd like to
conduct two-sided inference on R.
The standard workhorses for inference:
1. Hypothesis Testing: H0 : = 0 with p-value
p(0 , x) = inf {x (0 )}.
The Setup
Suppose you have data X from some model P and youd like to
conduct two-sided inference on R.
The standard workhorses for inference:
1. Hypothesis Testing: H0 : = 0 with p-value
p(0 , x) = inf {x (0 )}.
(1 - ) CI
b
p(B , x) p(I , x) ?
An example
An example
i.i.d.
An example
i.i.d.
Where X = 1.
Y X
N(0, 1)
1 + 2
Y X
N(0, 1)
1 + 2
and thus V is a pivotal quantity.
V =
Y X
N(0, 1)
1 + 2
and thus V is a pivotal quantity. We can use this to construct a
V =
Thank You!