Numeric variable read as nominal (PASW 18)
Locked
Numeric variable read as nominal (PASW 18)
 
|
Just installed PASW 18 and am running some
SPSS 15 tutorials through to see if they work and to check differences in
displays. On the very first run it has just read in a numeric variable
(number of people in household, F2.0) but the Data Editor displays the type as
Nominal, not Scale. Age last birthday (F2.0) displays as
Scale.
|
Re: Numeric variable read as nominal (PASW 18)
|
When a data source does not already have a measurement level recorded, the level is actually set to unknown (which is masked in the Data Editor). After the first data pass, unknown levels are set based on a heuristic that you can control in Edit>Options>Data Reading External Data. Old sav files might not have a level recorded. You can use Data>Define Variable Properties to apply a better set of heuristics to guess a level or, of course, you can set the level in that dialog, in the DE, or via VARIABLE LEVEL syntax. Once the file is saved, the chosen level is recorded. HTH, Jon Peck SPSS, an IBM Company [hidden email] 312-651-3435
Just installed PASW 18 and am running some SPSS 15 tutorials through to see if they work and to check differences in displays. On the very first run it has just read in a numeric variable (number of people in household, F2.0) but the Data Editor displays the type as Nominal, not Scale. Age last birthday (F2.0) displays as Scale. |
Re: Numeric variable read as nominal (PASW 18)
|
|
Strange: even with syntax...?
DATA LIST FILE =
'e:bsa86.dat'
RECORDS = 23 /15 v1508 8-9 v1512 12-13. Also, the format of syntax copied from elsewhere is
changed in the syntax file.
However, I really like the automatic colour coding
of commands, subcommands and specifications,
but it didn't pick up /HBAR, just left it
black. I use colour coding in my tutorials, but if your convention remains
stable remains stable, I'll take time to change all mine to match (bit of a
chore, but avoids confusion in students and trainees).
|
Re: Numeric variable read as nominal (PASW 18)
|
|
Jon
Should have said that, in the Data Editor, N of
people in household displays as Numeric, Width 2, Decimals 0, but as Measure
Nominal. Can this be right? If I want my readers to start using
SPSS, and stay with it, all confusion must be removed. As my old colleague
and computer wizard Jim Ring used to say, "If it's Hall-proof, it's idiot
proof!" Do they have Test to destruction in the USA?
Obsessed with minutiae (nit-picker?) I may be, but If there's even the
slightest error or confusion anywhere in the bits I use, be sure I'll find it
eventually.
Cheers
John
From: [hidden email]
|
Re: Numeric variable read as nominal (PASW 18)
|
|
In reply to this post by John F Hall
This was the command, but the line-wrapping excuse
doesn't wash here: I had to chanfge the format again. For ingénue
students, everything has to work as expected. Also my tutorials move
quickly to the use of abbreviated syntax to save time (and key
depressions). GUI must not be the tail that wags the dog!
Here's the syntax with /HBAR you asked about
(unabbreviated syntax)
FREQUENCIES
VARIABLES =
V1508
/HBAR /STATISTICS MEAN MEDIAN MODE.
|
Re: Numeric variable read as nominal (PASW 18)
|
|
In reply to this post by John F Hall
Jon Peck SPSS, an IBM Company [hidden email] 312-651-3435
Jon Should have said that, in the Data Editor, N of people in household displays as Numeric, Width 2, Decimals 0, but as Measure Nominal. Can this be right? >I don't see any contradiction here. Type and format are unrelated to measurement level, although scale-level strings would be rather weird. If I want my readers to start using SPSS, and stay with it, all confusion must be removed. As my old colleague and computer wizard Jim Ring used to say, "If it's Hall-proof, it's idiot proof!" Do they have Test to destruction in the USA? Obsessed with minutiae (nit-picker?) I may be, but If there's even the slightest error or confusion anywhere in the bits I use, be sure I'll find it eventually. Cheers John From: Jon K Peck To: John F Hall Sent: Monday, March 22, 2010 8:11 PM Subject: Re: [SPSSX-L] Numeric variable read as nominal (PASW 18) see below. Jon Peck SPSS, an IBM Company peck@... 312-651-3435
Strange: even with syntax...? >>>Pretty much everything is actually done with syntax even in the few cases where it is not visible. ML settings have been in place for a long time, though. DATA LIST FILE = 'e:bsa86.dat' RECORDS = 23 /15 v1508 8-9 v1512 12-13. Also, the format of syntax copied from elsewhere is changed in the syntax file. >>>>Should only affect line wrapping However, I really like the automatic colour coding of commands, subcommands and specifications, but it didn't pick up /HBAR, just left it black. I use colour coding in my tutorials, but if your convention remains stable remains stable, I'll take time to change all mine to match (bit of a chore, but avoids confusion in students and trainees). >>>What was the full command that did not highlight HBAR? Offhand, I don't recall a subcommand by that name - and neither did the syntax editor, which is why it left the text black. You will also see that abbreviated commands are not recognized: it was just too hard to get all the abbreviation resolution to work in the syntax editor, and we now discourage the use of abbreviations for various reasons. As for the colors for syntax coloring, these are settable via Edit>Options>Syntax Editor. In our documentation, though we generally show things with the default settings. ----- Original Message ----- From: Jon K Peck To: [hidden email] Sent: Monday, March 22, 2010 6:42 PM Subject: Re: Numeric variable read as nominal (PASW 18) When a data source does not already have a measurement level recorded, the level is actually set to unknown (which is masked in the Data Editor). After the first data pass, unknown levels are set based on a heuristic that you can control in Edit>Options>Data Reading External Data. Old sav files might not have a level recorded. You can use Data>Define Variable Properties to apply a better set of heuristics to guess a level or, of course, you can set the level in that dialog, in the DE, or via VARIABLE LEVEL syntax. Once the file is saved, the chosen level is recorded. HTH, Jon Peck SPSS, an IBM Company [hidden email] 312-651-3435
Just installed PASW 18 and am running some SPSS 15 tutorials through to see if they work and to check differences in displays. On the very first run it has just read in a numeric variable (number of people in household, F2.0) but the Data Editor displays the type as Nominal, not Scale. Age last birthday (F2.0) displays as Scale. |
Re: Numeric variable read as nominal (PASW 18)
|
|
In reply to this post by John F Hall
Jon Peck SPSS, an IBM Company [hidden email] 312-651-3435
This was the command, but the line-wrapping excuse doesn't wash here: I had to chanfge the format again. >>>Send me a file that get reformatted. For ingénue students, everything has to work as expected. Also my tutorials move quickly to the use of abbreviated syntax to save time (and key depressions). GUI must not be the tail that wags the dog! Here's the syntax with /HBAR you asked about (unabbreviated syntax) FREQUENCIES VARIABLES = V1508 /HBAR /STATISTICS MEAN MEDIAN MODE. HBAR is nowhere documented and must have been superseded many years ago, although it is still honored. FREQ currently documents /BARCHART /PIECHART, and /HISTOGRAM graphics subcommands each with several parameters. I vaguely recall reworking that command when we added more graphics choices. It was certainly undocumented in V15. ----- Original Message ----- From: Jon K Peck To: John F Hall Sent: Monday, March 22, 2010 8:11 PM Subject: Re: [SPSSX-L] Numeric variable read as nominal (PASW 18) see below. Jon Peck SPSS, an IBM Company peck@... 312-651-3435
Strange: even with syntax...? >>>Pretty much everything is actually done with syntax even in the few cases where it is not visible. ML settings have been in place for a long time, though. DATA LIST FILE = 'e:bsa86.dat' RECORDS = 23 /15 v1508 8-9 v1512 12-13. Also, the format of syntax copied from elsewhere is changed in the syntax file. >>>>Should only affect line wrapping However, I really like the automatic colour coding of commands, subcommands and specifications, but it didn't pick up /HBAR, just left it black. I use colour coding in my tutorials, but if your convention remains stable remains stable, I'll take time to change all mine to match (bit of a chore, but avoids confusion in students and trainees). >>>What was the full command that did not highlight HBAR? Offhand, I don't recall a subcommand by that name - and neither did the syntax editor, which is why it left the text black. You will also see that abbreviated commands are not recognized: it was just too hard to get all the abbreviation resolution to work in the syntax editor, and we now discourage the use of abbreviations for various reasons. As for the colors for syntax coloring, these are settable via Edit>Options>Syntax Editor. In our documentation, though we generally show things with the default settings. ----- Original Message ----- From: Jon K Peck To: [hidden email] Sent: Monday, March 22, 2010 6:42 PM Subject: Re: Numeric variable read as nominal (PASW 18) When a data source does not already have a measurement level recorded, the level is actually set to unknown (which is masked in the Data Editor). After the first data pass, unknown levels are set based on a heuristic that you can control in Edit>Options>Data Reading External Data. Old sav files might not have a level recorded. You can use Data>Define Variable Properties to apply a better set of heuristics to guess a level or, of course, you can set the level in that dialog, in the DE, or via VARIABLE LEVEL syntax. Once the file is saved, the chosen level is recorded. HTH, Jon Peck SPSS, an IBM Company [hidden email] 312-651-3435
Just installed PASW 18 and am running some SPSS 15 tutorials through to see if they work and to check differences in displays. On the very first run it has just read in a numeric variable (number of people in household, F2.0) but the Data Editor displays the type as Nominal, not Scale. Age last birthday (F2.0) displays as Scale. |
Re: Numeric variable read as nominal (PASW 18)
|
|
In reply to this post by Jon K Peck
This is an example of the limitations of any rules you could apply to automatically assign measurement level. Numeric codes are frequently used to represent categorical data. The relatively simple rule that is applied when reading external data sources (or very old SAV files) is based on the number of unique values recorded for each variable. The default is 24. So a numeric variable with less than 24 unique values is classified as nominal, and a variable with 24 or more unique values is classified as scale. Number of people in the household is clearly scale, but it's a variable that has only a few unique values, so it gets automatically classified as nominal. Slightly more complex rules are in the works for the next release, but this particular example would still probably be classified as scale. Note that there is no rule for automatically classifying a numeric variable as ordinal. All string variables are automatically classified as nominal.
Jon Peck SPSS, an IBM Company [hidden email] 312-651-3435
Jon Should have said that, in the Data Editor, N of people in household displays as Numeric, Width 2, Decimals 0, but as Measure Nominal. Can this be right? >I don't see any contradiction here. Type and format are unrelated to measurement level, although scale-level strings would be rather weird. If I want my readers to start using SPSS, and stay with it, all confusion must be removed. As my old colleague and computer wizard Jim Ring used to say, "If it's Hall-proof, it's idiot proof!" Do they have Test to destruction in the USA? Obsessed with minutiae (nit-picker?) I may be, but If there's even the slightest error or confusion anywhere in the bits I use, be sure I'll find it eventually. Cheers John From: Jon K Peck To: John F Hall Sent: Monday, March 22, 2010 8:11 PM Subject: Re: [SPSSX-L] Numeric variable read as nominal (PASW 18) see below. Jon Peck SPSS, an IBM Company peck@... 312-651-3435
Strange: even with syntax...? >>>Pretty much everything is actually done with syntax even in the few cases where it is not visible. ML settings have been in place for a long time, though. DATA LIST FILE = 'e:bsa86.dat' RECORDS = 23 /15 v1508 8-9 v1512 12-13. Also, the format of syntax copied from elsewhere is changed in the syntax file. >>>>Should only affect line wrapping However, I really like the automatic colour coding of commands, subcommands and specifications, but it didn't pick up /HBAR, just left it black. I use colour coding in my tutorials, but if your convention remains stable remains stable, I'll take time to change all mine to match (bit of a chore, but avoids confusion in students and trainees). >>>What was the full command that did not highlight HBAR? Offhand, I don't recall a subcommand by that name - and neither did the syntax editor, which is why it left the text black. You will also see that abbreviated commands are not recognized: it was just too hard to get all the abbreviation resolution to work in the syntax editor, and we now discourage the use of abbreviations for various reasons. As for the colors for syntax coloring, these are settable via Edit>Options>Syntax Editor. In our documentation, though we generally show things with the default settings. ----- Original Message ----- From: Jon K Peck To: [hidden email] Sent: Monday, March 22, 2010 6:42 PM Subject: Re: Numeric variable read as nominal (PASW 18) When a data source does not already have a measurement level recorded, the level is actually set to unknown (which is masked in the Data Editor). After the first data pass, unknown levels are set based on a heuristic that you can control in Edit>Options>Data Reading External Data. Old sav files might not have a level recorded. You can use Data>Define Variable Properties to apply a better set of heuristics to guess a level or, of course, you can set the level in that dialog, in the DE, or via VARIABLE LEVEL syntax. Once the file is saved, the chosen level is recorded. HTH, Jon Peck SPSS, an IBM Company [hidden email] 312-651-3435
Just installed PASW 18 and am running some SPSS 15 tutorials through to see if they work and to check differences in displays. On the very first run it has just read in a numeric variable (number of people in household, F2.0) but the Data Editor displays the type as Nominal, not Scale. Age last birthday (F2.0) displays as Scale. |
Re: Numeric variable read as nominal (PASW 18)
|
|
Thanks for the explanation. I tested the same
syntax on the same data with both 15 and 18.
DATA
LIST FILE =
'e:bsa86.dat'
RECORDS = 23 /15 v1508 8-9 . VARIABLE LABELS V1508 'Q105a Persons in household' . EXECUTE . Data Editors displayed:
SPSS15:
v1508 Numeric 2 0 Q105a Persons in
household None None 7 Right Scale
PASW18:
v1508 Numeric 2 0 Q105a Persons in
household None None 7 Right Nominal
Bit confusing when students have just been taught
levels of measurement and this is their first exercise!
On another point, I thought 18 might be different when it opened in Data
View the first time I used it. However, on opening and closing both 15 and
18 I discovered that when you open SPSS, it remembers whether your last
Data Editor was in Data View or Variable View. Tutors need to
know these things and point them out to nervous beginners. It would be
nice if it could also store the column widths in Variable View once
they have been adjusted.
----- Original Message -----
|
Automatic reply: Numeric variable read as nominal (PASW 18)
|
|
I will be out of the office from March 22-26th, returning on the 29th. I will be checking e-mail on a limited basis. If you need immediate assistance please contact me at 503/539-5606. Thank you. Kelly
===================== To manage your subscription to SPSSX-L, send a message to [hidden email] (not to SPSSX-L), with no body text except the command. To leave the list, send the command SIGNOFF SPSSX-L For a list of commands to manage subscriptions, send the command INFO REFCARD |
«
Return to SPSSX Discussion
|
1 view|%1 views
| Free forum by Nabble | Edit this page |