[R] how to get values within a threshold

[William Dunlap]

You may want to append -Inf (or 0 if you know the data cannot be
negative) to the start of your 'values' vector so you don't
have to write code to catch the cases when a threshold is below
the range of the values.
   > findInterval(thresholds, c(0,values,Inf))
   [1] 1 5 5 5 8
   > c(0, values, Inf)[.Last.value]
   [1] 0.000 1.874 1.874 1.874 7.722

Bill Dunlap
Spotfire, TIBCO Software
wdunlap tibco.com


> -----Original Message-----
> From: Zhang Weiwu [mailto:zhangweiwu at realss.com]
> Sent: Friday, September 13, 2013 8:52 AM
> To: William Dunlap; smartpink111 at yahoo.com
> Cc: r-help at r-project.org
> Subject: RE: [R] how to get values within a threshold
> 
> 
> 
> On Fri, 13 Sep 2013, William Dunlap wrote:
> 
> >> findInterval(thresholds, values)
> > [1] 1 4 4 4 7
> 
> Thanks a lot! But now I have a new problem, a typical R issue perhaps.
> 
> First, let's look at  a successful case:
> 
>  	> thresholds <- c(1,3,5,7,9)
>  	> values <- c(0.854, 1.648, 1.829, 1.874, 7.670, 7.673, 7.722)
>  	> values[findInterval(thresholds, values)]
>  	[1] 0.854 1.874 1.874 1.874 7.722
> 
> Then a new batch of values came, notice only the first element of new values
> differ:
> 
>  	> thresholds <- c(1,3,5,7,9)
>  	> values <- c(1.254, 1.648, 1.829, 1.874, 7.670, 7.673, 7.722)
>  	> findInterval(thresholds, values)
>  	[1] 0 4 4 4 7
>  	> values[findInterval(thresholds, values)]
>  	[1] 1.874 1.874 1.874 7.722
> 
> This is a surprise. The desirable output is:
> 
>  	[1] 0     1.874 1.874 1.874 7.722
> 
> This is desirable, because so maintains the same number of elements during
> calculation. (You may suggest leaving out the indices and try to calculate
> maximum-values-below-threshold directly, but the indices are useful to
> address other fields in the data frame whence values came.)
> 
> This problem can be simplified as following:
> 
> in R, we have:
>  	> a <- 1:10
>  	> a[c(1,3)]
>  	[1] 1 3
>  	> a[c(0,3)]
>  	[1] 3
> 
> While I was hoping to get:
>  	> a <- 1:10
>  	> a[c(1,3)]
>  	[1] 1 3
>  	> a[c(0,3)]
>  	[1] 0 3
> 
> The straightforward solution, is to shift the whole test values one
> position, so that the first value is always zero:
> 
>  	> values <- c(0, 1.254, 1.648, 1.829, 1.874, 7.670, 7.673, 7.722)
> 
> This solution, despite begetting a train of changes elsewhere in the code,
> is semantically wrong, since the first element of values should be the first
> value, now it is actually the 0-th value.
> 
> What would you do in the case?