Hi! I’ve been inactive on this list for a long time but am now trying to revive my memory. I’ve got a question from my son who is a SAS-programmer. He got a SPSS-file with an age variable (probably the result of translation from another data base system such as DBMSCOPY or similar). There he got a frequency table of the age variable with 2 occurences of 0 (zero). I remember having this problem a long time ago but do not remember the explanation for it. Can anybody help? best Staffan Lindberg Sweden |
The values are actually different but the formatting makes them appear to be the same. Increase the number of decimals in the variable format to see the different values. On Tue, Feb 6, 2018 at 10:31 AM Staffan Lindberg <[hidden email]> wrote:
-- |
Jon, I'm pretty sure that you have nailed it, number of decimals.
The question does remind me that I used a computer system once, which
distinguished between "zero" and "minus-zero"; and I ran a program which assumed Fortran would read blanks as minus-zero for Missing and then tested for Missing. My available Fortran did not implement "read" that way, so I had to
re-punch the data cards to replace blanks with "-0" in order to designate the Missing.
I think they don't make computers like that any more.
-- Rich Ulrich From: SPSSX(r) Discussion <[hidden email]> on behalf of Jon Peck <[hidden email]>
Sent: Tuesday, February 6, 2018 12:34:16 PM To: [hidden email] Subject: Re: Double occurence of 0 (Zero) The values are actually different but the formatting makes them appear to be the same. Increase the number of decimals in the variable format to see the different values.
On Tue, Feb 6, 2018 at 10:31 AM Staffan Lindberg <[hidden email]> wrote:
--
|
I remember such a system. -- long extinct. But in such a system, Statistics would still consider those two values to be the same, I think, and put the count in a single row, as long as it was not a string variable. On Tue, Feb 6, 2018 at 11:11 AM Rich Ulrich <[hidden email]> wrote:
-- |
In reply to this post by Staffan Lindberg
The standard internal SPSS format for a number is single- or double-precision floating point,
regardless of what SPSS or another package says to use as the format to display it.
Use the E-format to show all decimals - whatever the documentation says is the maximum.
Ordinary, rational fractions are stored imprecisely unless they represent a power of 2, like 1/2, 1/4, ....
An age of one day, transformed to "1/365 years", will look like zero; that is usually fixed by using
Truncate( ) or Round( ) when doing the Compute.
Trickier to spot - subtracting fractions can give something very /near/ to an integer, but not exactly. Thus, if you subtract (1/3+1/3) from 2/3, you may or may not get zero, depending on the exact
internal representations. The difference will be last couple of reportable digits.
-- Rich Ulrich
From: SPSSX(r) Discussion <[hidden email]> on behalf of Staffan Lindberg <[hidden email]>
Sent: Tuesday, February 6, 2018 12:31:47 PM To: [hidden email] Subject: Double occurence of 0 (Zero) Hi!
I’ve been inactive on this list for a long time but am now trying to revive my memory. I’ve got a question from my son who is a SAS-programmer. He got a SPSS-file with an age variable (probably the result of translation from another data base system such as DBMSCOPY or similar). There he got a frequency table of the age variable with 2 occurences of 0 (zero). I remember having this problem a long time ago but do not remember the explanation for it.
Can anybody help?
best
Staffan Lindberg Sweden |
Free forum by Nabble | Edit this page |