 Notice: On April 23, 2014, Statalist moved from an email list to a forum, based at statalist.org.

# st: RE: How to improve accuracy in numerical integrations using Stata

 From "Feiveson, Alan H. (JSC-SK311)" To "statalist@hsphsun2.harvard.edu" Subject st: RE: How to improve accuracy in numerical integrations using Stata Date Fri, 23 Dec 2011 11:03:42 -0600

```Tiago - You can also consider Gaussian integration - for example see my presentation at the 2004 Stata users' group meeting.

http://www.stata.com/meeting/3nasug/abstracts.html

Al Feiveson

-----Original Message-----
From: owner-statalist@hsphsun2.harvard.edu [mailto:owner-statalist@hsphsun2.harvard.edu] On Behalf Of Tiago V. Pereira
Sent: Friday, December 23, 2011 10:45 AM
To: statalist@hsphsun2.harvard.edu
Subject: st: How to improve accuracy in numerical integrations using Stata

Dear Maarten,

I don't have a clue on how to select specific knots, but I will
investigate that.
As usual (since 2005!) I thank you very much for you time and extremely

All the best,

Tiago

---
Dear statalisters,

I am using -integ- to numerically integrate a set of functions.

An example of a function to integrate:

function_y  = -2*normal((-`x'-`r'*z)/sqrt(1-`r'^2))*normalden(z)

for variable z.

In my case, the domain ranges from 0 to `x'.

So, what I am doing is  the following:

*/ ------------ start example --------------
local r = 0.1
local x = 6
drop _all
range z 0 `x' 1000
generate y  = -2*normal((-`x'-`r'*z)/sqrt(1-`r'^2))*normalden(z)
dydx     y z, gen(yprime)
integ    y z, gen(Sy)
*/ ------- end example ------------------

dis r(integral)
-5.288096*07214*e-10

to gain more precision, I have manually edited -integ- to compute values
using the double format (i.e. instead of 'gen float variable = ', it is
using 'gen double variable =').

It seems that some precision is gained:

[using the exactly same code above, but using the 'double' version, one
gets:]

dis r(integral)

-5.288096*31782*e-10

I know that the correct answer would be something like:

-5.28809630924643245856745711e-10

which is obtained from numerical integration using a C program (supposed
to be the most precise approach I am aware of).

Do you have any ideas on how to further increase the precision for
numerical integration in Stata? The problem is that I am working on heavy
tails (alpha levels below 10^-8).

All the best,

Tiago

*
*   For searches and help try:
*   http://www.stata.com/help.cgi?search
*   http://www.stata.com/support/statalist/faq
*   http://www.ats.ucla.edu/stat/stata/

*
*   For searches and help try:
*   http://www.stata.com/help.cgi?search
*   http://www.stata.com/support/statalist/faq
*   http://www.ats.ucla.edu/stat/stata/
```