R help - Oct 2009 - Cannot calculate mean() for double vector

I've been using R recently to analyze some data, but I'm having a
problem using the mean() function.

I imported the original data set as a vector of integers, x and then
calculated a exponential moving average of the data, x_ema. This part
worked fine.

Then, I tried to find the mean squared error between the original
series and the moving average, using mse = mean((x - x_ema)^2). This
gives N/A as a result, which seems to be the result of the mean
function. When I run mean() on x_ema, which is of data type double, it
always returns N/A. I can find the mean of the original integer data
just fine, as well as for a simple test vector of 10 doubles, but it
never seems to return a usable result for the x_ema vector. Is this
because the vector is too long? (Both x and x_ema contain around 40000
values). Am I running into some memory limit? I can do other
calculations with x_ema just fine, it's only the mean() function that
doesn't seem to work. All the information in the help seems to
indicate that this operation should work without a problem.

If it matters at all, I'm using R 2.9.2 running on Windows Vista.

I'd appreciate any help or suggestions as to what might be wrong.

Thank you,
Reuben Bellika

Have you tried:
mean(x)
mean(as.numeric(as.character(x)))
mean(x_ema)
mean(as.numeric(as.character(x_ema)))

What is the result of the following:
which(is.na(as.numeric(as.character(x_ema))))

Abit hard since you don't provide the data, but there may be an NA or
character value that is causing the error.

Hope this helps at a bit. 



--- On Fri, 10/16/09, Reuben Bellika <rubenyi at gmail.com> wrote:
> From: Reuben Bellika <rubenyi at gmail.com>
> Subject: [R] Cannot calculate mean() for double vector
> To: r-help at r-project.org
> Date: Friday, October 16, 2009, 12:06 PM
> I've been using R recently to analyze
> some data, but I'm having a
> problem using the mean() function.
> 
> I imported the original data set as a vector of integers, x
> and then
> calculated a exponential moving average of the data, x_ema.
> This part
> worked fine.
> 
> Then, I tried to find the mean squared error between the
> original
> series and the moving average, using mse = mean((x -
> x_ema)^2). This
> gives N/A as a result, which seems to be the result of the
> mean
> function. When I run mean() on x_ema, which is of data type
> double, it
> always returns N/A. I can find the mean of the original
> integer data
> just fine, as well as for a simple test vector of 10
> doubles, but it
> never seems to return a usable result for the x_ema vector.
> Is this
> because the vector is too long? (Both x and x_ema contain
> around 40000
> values). Am I running into some memory limit? I can do
> other
> calculations with x_ema just fine, it's only the mean()
> function that
> doesn't seem to work. All the information in the help seems
> to
> indicate that this operation should work without a
> problem.
> 
> If it matters at all, I'm using R 2.9.2 running on Windows
> Vista.
> 
> I'd appreciate any help or suggestions as to what might be
> wrong.
> 
> Thank you,
> Reuben Bellika
> 
> ______________________________________________
> R-help at r-project.org
> mailing list
> https://stat.ethz.ch/mailman/listinfo/r-help
> PLEASE do read the posting guide
http://www.R-project.org/posting-guide.html
> and provide commented, minimal, self-contained,
> reproducible code.
>

Reuben Bellika wrote:
>
> I've been using R recently to analyze some data, but I'm having a
> problem using the mean() function.
> 
> I imported the original data set as a vector of integers, x and then
> calculated a exponential moving average of the data, x_ema. This part
> worked fine.
> 
> Then, I tried to find the mean squared error between the original
> series and the moving average, using mse = mean((x - x_ema)^2). This
> gives N/A as a result, which seems to be the result of the mean
> function. When I run mean() on x_ema, which is of data type double,
>  it always returns N/A. 
>
So, x_ema includes one (or more) NA (and not N/A) in it.

Test: if (any(is.na(x_ema))) cat("Oops! NAs in x_ema\n")

If you want to get which of them are na: which(is.na(x_ema))

Alberto Monteiro

OK. It looks like I just have several NA values at the start of my array:
> which (is.na(x_ema))
[1] 1 2 3 4 5 6 7 8 9

That make sense, because the moving average is not defined for those
positions. I'll just have to set those values to zero:
> x_ema = replace(x_ema, which(is.na(x_ema)), 0)
> which (is.na(x_ema))
integer(0)

The mean() call works now and I can get on with my work. I'll have to
remember to condition the data like this in the future.

Thanks for the help!

Reuben Bellika

It looks like I should read the documentation a bit more carefully.
Simply ignoring the NA values when calculating the mean *is* a much
better solution. (Although having 10 values out of a vector of 40000
or so set to zero is *not* going to bias the mean toward zero in any
significant way.)

Thanks to everyone for your help.

Reuben