Bookmark and Share

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


[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: st: trying to round values and store in a macro, problem with double/floating precision number storage


From   Nick Cox <[email protected]>
To   [email protected]
Subject   Re: st: trying to round values and store in a macro, problem with double/floating precision number storage
Date   Thu, 5 May 2011 19:19:11 +0100

Maarten gave positive advice but some pitfalls here deserve more explanation.

First, although -round(,)- allows non-integer second arguments, most
are useless or in practice misused.

In fact I would say that the Stata documentation is very misleading on
this point. Tthe help for -round()- gives as an example (in Stata
10.1; I am away from my main machine)

"The rounding definition is generalized for y != 1.  With y = .01, for
instance, x is rounded to two decimal places; round(sqrt(2),.01) is
1.41."

This just perpetuates a fallacy that is elsewhere painfully explained.
There can be no sense in which sqrt(2) _is_ 1.41 as such a number
cannot be held exactly in Stata or indeed any machine based on binary
arithmetic.

It is true that e.g.

. di round(sqrt(2),.01)
1.41

but that is at least in part because the default display format works
as you want here.

When people want to show a given number of decimal places, there is
only one way to do it directly and reliably: use an appropriate
display format. -round()- works poorly if at all for such purposes.

Second, putting a numeric constant into a variable and then using
-summarize- to get it out again is unnecessary in two senses. Maarten
gave alternative code, but even setting aside the misuse of -round()-
these seven lines can be reduced to two:

               local mn`v'_snh`i' = round(r(mean),0.01);
               local std`v'_snh`i' = round(r(sd),0.01);

On Thu, May 5, 2011 at 4:44 PM, Woolton Lee <[email protected]> wrote:

[snip]
               gen float mn = round(r(mean),0.01);
               gen float sd = round(r(sd),0.01);
               sum mn;
               local mn`v'_snh`i'= r(mean);
               sum sd;
               local std`v'_snh`i' = r(sd);
               drop mn sd;

[snip]

> This loop creates the variables mn and sd and rounds the result of the
> numbers I want to two decimal places then uses summarize to store
> these values into macros.  I wonder if there is another more efficient
> way to do this?  I originally was doing all of the computations with
> only macros but I found that I could not round to two decimals without
> running into problems where a number like 55.87000000000000000001
> would show up in my tables.  Is there anyway to make this work doing
> the computations with just macros?

*
*   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/


© Copyright 1996–2018 StataCorp LLC   |   Terms of use   |   Privacy   |   Contact us   |   Site index