Message-ID: <loom.20090327T151919-812@post.gmane.org>
Date: 2009-03-27T15:28:42Z
From: Mark Lyman
Subject: Weighting data with normal distribution
Alice Lin <alice.ly <at> gmail.com> writes:
>
>
> I have a vector of binary data ? a string of 0?s and 1?s.
> I want to weight these inputs with a normal kernel centered around entry x
> so it is transformed into a new vector of data that takes into account the
> values of the entries around it (weighting them more heavily if they are
> near).
>
> Example:
> -
> - -
> - -
> 0 1 0 0 1 0 0 1 1 1 1
> If x = 3, it?s current value is 0 but it?s new value with the Gaussian
> weighting around would be something like .1*0+.5*1+1*0+0.5*0+.1*1= 0.6
>
> I want to be able to play with adjusting the variance to different values as
> well.
> I?ve found wkde in the mixtools library and think it may be useful but I
> have not figured out how to use it yet.
>
> Any tips would be appreciated.
>
> Thanks!
>
I don't know anything about wkde. But the filter function in stats package
should do what you want.
> x <- c(0, 1, 0, 0, 1, 0, 0, 1, 1, 1, 1)
> filter(x, c(.1, .5, 1, .5, .1))
Time Series:
Start = 1
End = 11
Frequency = 1
[1] NA NA 0.6 0.6 1.0 0.6 0.7 1.6 2.1 NA NA
In the signal package, there is also a variety of windows, including the
gausswin function. However, the filter function in the signal package masks the
filter function from the stats package
> stats::filter(x, gausswin(5, 2.68))
Mark Lyman
Statistician, ATK Launch Systems