Re: Gaps in plotted graph, probably resulting from real values being miscomputed as complex
- To: mathgroup at smc.vnet.net
- Subject: [mg123508] Re: Gaps in plotted graph, probably resulting from real values being miscomputed as complex
- From: "Oleksandr Rasputinov" <oleksandr_rasputinov at hmamail.com>
- Date: Fri, 9 Dec 2011 05:57:19 -0500 (EST)
- Delivered-to: l-mathgroup@mail-archive0.wolfram.com
On Thu, 08 Dec 2011 10:25:11 -0000, jdm <james.d.mclaughlin at gmail.com>
wrote:
> Plot[Log[2,
> PDF[NormalDistribution[0, 1],
> InverseCDF[NormalDistribution[0, 1], (1 - 2^(-x))]]^2], {x, 1,
> 256}, PlotRange -> {-500, 0}]
>
> plots up to about x=56, and then nothing.
>
> Based on similar behaviour under another system, I suspect that some
> sort of
> calculation error as 1 - 2^(-x) approaches 1 is causing some of the
> values that should be plotted to be wrongly computed as complex
> numbers, but have no way of confirming this and don't understand why
> the real-valued values aren't being plotted for their corresponding x
> if this is the case.
>
> Can anyone suggest anything?
>
> Thanks,
>
> James McLaughlin.
>
Working in machine precision without precision tracking leads to numerical
errors in this case. Switching on precision tracking fixes the issue:
Plot[
Log[2,
PDF[
NormalDistribution[0, 1],
InverseCDF[NormalDistribution[0, 1], (1 - 2^(-x))]
]^2
], {x, 1, 256},
PlotRange -> {-500, 0}, WorkingPrecision -> $MachinePrecision
]