Skip to content

problem with Gauss Hermite ( x and w )

6 messages · casperyc, Rui Barradas, R. Michael Weylandt

#
Hi all,

I am using the 'gaussHermite' function from the 'pracma' library

############ CODES ###########
library(pracma)
cc=gaussHermite(10)
cc$x^2
cc$x^5
cc$x^4
############ CODES ###########

as far so good. However, it does NOT work for any NON integer values, say

############ CODES ###########
cc$x^(2.5)
cc$x^(-2.5)
############ CODES ###########

But just think about it numberically, it should work.....

why this is the case?

Is there a reason for getting the "NaN"s?

Thanks!







-----
######################
PhD candidate in Statistics
Big R Fan
Big LEGO Fan
Big sTaTs Fan
######################

--
View this message in context: http://r.789695.n4.nabble.com/problem-with-Gauss-Hermite-x-and-w-tp4622115.html
Sent from the R help mailing list archive at Nabble.com.
#
Taking negative numbers to fractional powers gives NaNs.... that's
just how it works.

Unless you want to use complex numbers (which R does not by default):

as.complex(cc$x) ^ (2.5)

Michael
On Wed, May 9, 2012 at 7:22 PM, casperyc <casperyc at hotmail.co.uk> wrote:
#
Hi,

 I know what complex number are, but I am not sure what you meant by that?


##############CODES###########
[1] 0.1109032
[1] -0.1109032
##############CODES###########

works fine.

Negative powers mean they take the reciprocal and as far as I am concerned,
real^real is just a real number.

Am I mistaking something basic?

Thanks.

Casper

-----
######################
PhD candidate in Statistics
Big R Fan
Big LEGO Fan
Big sTaTs Fan
######################

--
View this message in context: http://r.789695.n4.nabble.com/problem-with-Gauss-Hermite-x-and-w-tp4622115p4624395.html
Sent from the R help mailing list archive at Nabble.com.
#
Hello,
Yes, you are.

real^real is not necessarily real.

The most well known example is (-1)^0.5 = imaginary unit.
When you say that -2.5^(-2.4) is real you're computing the negative power of
a POSITIVE real, 2.5 then taking the result's  symmetric.

(-2.5)^(-2.4)
[1] NaN

Or
(-1)^0.5
[1] NaN
-1^0.5
[1] -1

(-1 + 0i)^0.5 # algebraically, equal to (-1)^0.5 above, but not in R
notation.
[1] 0+1i

Hope this helps,

Rui Barradas



casperyc wrote
--
View this message in context: http://r.789695.n4.nabble.com/problem-with-Gauss-Hermite-x-and-w-tp4622115p4624486.html
Sent from the R help mailing list archive at Nabble.com.
#
Rui Barradas wrote
Damn, can't believe it! It's a silly mistake!

Now that something wonders me is that when applying the Gaussian Hermit, 

sum w f(x_i)

What happens when f(x_i) does not work?
This has nothing to do with R, I will try read more on the material myself.

Thank you all!

Casper


-----
######################
PhD candidate in Statistics
Big R Fan
Big LEGO Fan
Big sTaTs Fan
######################

--
View this message in context: http://r.789695.n4.nabble.com/problem-with-Gauss-Hermite-x-and-w-tp4622115p4624557.html
Sent from the R help mailing list archive at Nabble.com.
#
Are you getting caught on order of operations? Note that unary minus
has lower precedence than exponentiation (as it does in math) so

-2.5^(-2.4)

is

x <- 2.5^(-2.4)
-x

Otherwise, I'm not at all sure what your question is: can you give an
example of what you think you should get (and how to get it) and what
R is giving you instead?

Michael
On Thu, May 10, 2012 at 2:30 PM, casperyc <casperyc at hotmail.co.uk> wrote: