Message Boards Message Boards

0
|
7104 Views
|
3 Replies
|
0 Total Likes
View groups...
Share
Share this post:

Using Nest for Convolution of n Uniform Distributions

Posted 10 years ago

I want to define the pdf function that is the convolution of n uniform [0,1] pdfs using the Nest function/process. I realize I can use a for-like loop to accomplish this, but I am trying to expand my horizons and understand Mathematica and programming better... I want to use Nest but cannot figure out what I am doing wrong (probably something simply, but I have spent 2+ hours trying to get below to work):

ClearAll[U, UConv];

U[x_] = PDF[UniformDistribution[], x];

Convolve[U[x], U[x], x, y]

UConv[w_] := {Convolve[U[x], w[[1]][x], x, w[[2]]], w[[2]]};

Nest[UConv, {U, y}, 2]

I was hoping/expecting that Nest[UConv, {U, y}, 2] gives the same result as Convolve[U[x], U[x], x, y], but it doesn't. I cannot even figure out how to interpret Nest[UConv, {U, y}, 2]'s result! Any hints, help, or suggestions would be greatly appreciated. ds

POSTED BY: David S
3 Replies
Posted 10 years ago

I think the correct definition should read

UConv[w_] := {Convolve[U[x], w[[1]] /. w[[2]] -> x, x, w[[2]]], w[[2]]}

in order for Nest to give the expected output (confer details above).

Best, Xavier

POSTED BY: Xavier Roy

Your code has some errors. I think you want:

UConv[w_] := {Convolve[U[x], w[[1]], x, w[[2]]], w[[2]]}

Nest[UConv, {U[x], y}, 2]

POSTED BY: Sean Clarke
Posted 10 years ago

Hi David,

With the Nest command one applies a function several times to an expression. The evaluation of, for instance,

Nest[U,x,2]

gives the same output as

U[U[x]]

So in your example

Nest[UConv, {U, y}, 2]

is equivalent to

UConv[UConv[{U, y}]]

With that in mind it should be easier to understand what is going on. The first element of Convolve in the output you obtain

{Convolve[Piecewise[{{1, 0 <= x <= 1}}, 0], Piecewise[{{2 - y, 1 < y < 2}, {y, 0 < y <= 1}}, 0][x], x, y], y}

is U[x] (confer definition of UConv), the second element is the first element of the output list given by

UConv[{U, y}]

to which an [x] is added on the right (confer again definition of UConv). The last element of Convolve and the last element of the output you obtain are both the second element of the output list given by

UConv[{U, y}]

so in this case y. Now, Mathematica does not evaluate

Convolve[Piecewise[{{1, 0 <= x <= 1}}, 0], Piecewise[{{2 - y, 1 < y < 2}, {y, 0 < y <= 1}}, 0][x], x, y]

because there is an extra [x] at the second Piecewise. You can check that the first piecewise function does not have any [x], because none is needed in effect. So the definition for UConv should be instead

UConv[w_] := {Convolve[U[x], w[[1]], x, w[[2]]], w[[2]]}

and the [x] should be written when calling it, namely

UConv[{U[x], y}]

However we run into a new problem when calling twice UConv, with

Nest[UConv,{U[x], y},2]

since the ouput is the same as

UConv[{U[x], y}]

(The second evaluation of UConv did not process.) The thing is that UConv[{U[x], y}] returns a Piecewise whose variable is y, but we want piecewise functions of variable x since this is what we ask for in the definition of UConv. This is the third argument of

UConv[w_] := {Convolve[U[x], w[[1]], x, w[[2]]], w[[2]]}

We simply need to change y (which is actually w[[2]]) to x by redefining UConv as

UConv[w_] := {Convolve[U[x], w[[1]] /. w[[2]] -> x, x, w[[2]]], w[[2]]}

Using this last definition, you should be able to Nest it without any problem (I have tried it with some examples and it seems to work OK) with

Nest[UConv, {U[x], y}, 2]

Note at last that, contrary to your remark, this is

UConv[{U[x], y}]

which is equivalent to

Convolve[U[x], U[x], x, y]

The evaluation of

Nest[UConv, {U[x], y}, 2]

is equivalent to

Convolve[U[x], U[x], x, y];
Convolve[U[y], %, y, t]

Best, Xavier

POSTED BY: Xavier Roy
Reply to this discussion
Community posts can be styled and formatted using the Markdown syntax.
Reply Preview
Attachments
Remove
or Discard

Group Abstract Group Abstract