[Date Prev][Date Next][Thread Prev][Thread Next][Date index][Thread index]
st: RE: question on datatypes
This is a frequent question on Statalist.
If you what are reading in, although it looks
like a standard numeric variable with a decimal
point, is really some kind of identifier or
code, then I think by far the safest strategy
is to read it in as a string variable and
then do string manipulations on it. This can
include removing the point.
Even if you read it in as -float- and then convert
to -long- it is all too likely that errors will
be introduced in the last digit or so.
As a thread yesterday advised, -search precision-
and look at the FAQs there mentioned. The underlying
issue is that computers are really doing calculations
in binary, not decimal. A massive amount of cleverness goes into
hiding this fact from the user, and even the programmer,
but it is inescapable and on occasions can bite you.
> I have variables that are coded in
> as 2-3 digits and 6 decimal places
> after that, e.g. 150.089873 and I want to convert this to get
> rid of the decimal places so it
> looks like 150089873 but when I just multiply by 1000000 then
> it keeps storing it as a "float"
> display i.e. 1.5E8 and more than that, the last decimal
> places are actually INCORRECT in the new
> version...it will say 1.50089875 instead...which is
> INCORRECT. I am using Stata 8 and that is
> where this problem occurs. I was wondering what the easiest
> way is to change datatypes that
> something is input in so that it will not be a problem.
* For searches and help try: