Search All of the Math Forum:

Views expressed in these public forums are not endorsed by NCTM or The Math Forum.

Notice: We are no longer accepting new posts, but the forums will continue to be readable.

Topic: dark halo around gaussian peak in Plot3D?
Replies: 9   Last Post: Mar 10, 2013 12:47 AM

 Messages: [ Previous | Next ]
 Bob Hanlon Posts: 906 Registered: 10/29/11
Re: dark halo around gaussian peak in Plot3D?
Posted: Mar 9, 2013 5:25 AM

Table[
Plot3D[
Evaluate[
PDF[BinormalDistribution[0], {x, y}]],
{x, -n, n}, {y, -n, n},
PlotRange -> All,
PlotPoints -> 50,
PlotLabel -> StringForm["n = ``", n],
ImageSize -> 300],
{n, {5, 10, 15, 20, 37, 40}}] //
Partition[#, 2] & // Grid

The problem starts at about 20 standard deviations and stops at about
37 standard deviations out on the tail. These are really small values.
Presumably, this is a precision issue that stops when the algorithms
just give up and call the results zero.

Table[
PDF[BinormalDistribution[0], {n, n}] // N,
{n, {20, 37, 40}}]

{3.0480870817634576*^-175, \
4.494427778507250610367056211787865772187`15.954589770191005*^-596,
2.1411598269883908369272155013252852`15.954589770191005*^-696}

However, even extreme WorkingPrecision does not appear to resolve the issue

With[{n = 40},
Plot3D[
Evaluate[PDF[BinormalDistribution[0], {x, y}]],
{x, -n, n}, {y, -n, n},
PlotRange -> All,
PlotPoints -> 50,
PlotLabel -> StringForm["n = ``", n],
ImageSize -> 300,
WorkingPrecision -> 2000,
PerformanceGoal -> "Quality"]]

Perhaps then the question is: why do you need results that far out on the tail?

Or you could work with the Log of the function

With[{n = 40},
Plot3D[
Evaluate[
Log[PDF[BinormalDistribution[0], {x, y}]] //
PowerExpand],
{x, -n, n}, {y, -n, n},
PlotRange -> All,
PlotPoints -> 50,
PlotLabel -> StringForm["n = ``", n],
ImageSize -> 600]]

Bob Hanlon

On Fri, Mar 8, 2013 at 11:06 AM, Michael B. Heaney <mheaney@alum.mit.edu> wrote:
> Here is simple example that demonstrates the problem:
>
> Plot3D[PDF[BinormalDistribution[0], {x, y}], {x, -40, 40}, {y, -40, 40},
>
> PlotRange -> All, PlotPoints -> 200]
>
> Thanks,
>
> Michael
>
>
> On Thu, Mar 7, 2013 at 8:49 PM, Bob Hanlon <hanlonr357@gmail.com> wrote:

>>
>> I don't see a "dark halo" with the following
>>
>> Plot3D[
>> PDF[BinormalDistribution[0], {x, y}],
>> {x, -3, 3}, {y, -3, 3}]
>>
>> Can you send a simple example that demonstrates the problem?
>>
>>
>> Bob Hanlon
>>
>>
>> On Thu, Mar 7, 2013 at 10:51 PM, Michael B. Heaney <mheaney@alum.mit.edu>
>> wrote:

>> >
>> > Hi,
>> >
>> > When I plot a gaussian peak in Plot3D, there is a dark halo around the
>> > peak. Why is this? How can I get rid of it?
>> >
>> > Thanks,
>> >
>> > Michael
>> >
>> > --
>> > ----------------------------------------------------------
>> > Michael B. Heaney
>> > 3182 Stelling Drive
>> > Palo Alto, CA 94303 USA
>> > mheaney@alum.mit.edu
>> > ----------------------------------------------------------
>> >
>> >

>
>
>
>
> --
> ----------------------------------------------------------
> Michael B. Heaney
> 3182 Stelling Drive
> Palo Alto, CA 94303 USA
> mheaney@alum.mit.edu
> ----------------------------------------------------------
>

Date Subject Author
3/8/13 Bob Hanlon
3/8/13 Michael B. Heaney
3/9/13 David Reiss
3/9/13 Bob Hanlon
3/9/13 Bob Hanlon
3/9/13 Michael B. Heaney
3/9/13 David Park
3/10/13 Michael B. Heaney