This is an important discussion
The traditional way to import a dataset in the "old time" was to import the data (mostly in numeric form) with a "Fortran-like" format and then add a dictionary with the "Var label" and he "Value label" set of instructions. In fact when importing from a .csv we are in the same tradition, with 2 files: data and dictionnary. And a traditional dictionnary with var label and value labels is not so difficult to organize.
In any case, I would prefer to stick with the numeric codes rather than use character values that are much more difficult to convert (the problem is the same when importing a "foreign file" in R coming from a .sav where it appears to be easier to use the numeric codes rather than the string coming from the values
The alternative proposed by Markus, to import from a ddi-xml structure is probably the most interesting as already a lot of tools allow to convert from survey tools to a ddi strcuture. In my memory the quexml developped in Australia was a step in this direction (https://quexml.acspri.org.au/) as well as a way to go directly from software like Limesurvey to SPSS (sorry PSPP) see https://quexmltools.acspri.org.au/
This is just a suggestion to explore this direction but this has to be analysed further
Dominique
Le 21.01.22 à 21:10, Elio Spinello a écrit :
For a codebook that doesn’t already exist, my suggestion would be to import from a spreadsheet or CSV file which has been formatted using specified column names for the variable name, label, type, etc.
From: Pspp-users <pspp-users-bounces+espinello=rpmconsulting.com@gnu.org> On Behalf Of 'Alan Mead'
Sent: Friday, January 21, 2022 11:10 AM
To: pspp-users@gnu.org
Subject: Re: Import Codebook
I agree. This is a good trick when you have an existing SAV file with the same codebook, but I don't see how this reduces the effort for the use case where you are importing 200 columns of unlabeled data?
-Alan
On 1/21/2022 12:54 PM, Elio Spinello wrote:
If memory serves me correctly, there is a Copy Data Properties tool that allows you to select another dataset or unopened SAV file and then copy the data properties from it into the active dataset.
Or you can copy and paste portions of the datasheet from one dataset to another.
https://www.youtube.com/watch?v=ZakhRd4aDAQ
I would think that one of those approaches would probably be the easiest to work with for both developers and users.
Elio Spinello
Elio Spinello, EdD
RPM Consulting, LLC
27943 Seco Canyon Rd #320
Santa Clarita, CA 91350-3872
Office: 818-831-7607
Cell: 818-570-3546
From: Pspp-users <pspp-users-bounces+espinello=rpmconsulting.com@gnu.org> On Behalf Of Ben Pfaff
Sent: Friday, January 21, 2022 10:29 AM
To: Alan Mead <amead@alanmead.org>
Cc: pspp-users <pspp-users@gnu.org>
Subject: Re: Import Codebook
If PSPP were to add a feature to import a codebook, what format should it be able to import it from?
On Fri, Jan 21, 2022 at 10:20 AM <amead@alanmead.org> wrote:
Yes, but variable labels aren't always that big a deal; value labels can be more critical. You should rename/label, but it's fairly easy to remember that V3 is sex. Good luck, however, remembering what the five responses 1, 2, 3, 4, 5 mean...
Elio ninja'd me last night because I spent a few minutes googling whether there was a way to import a code book. I don't think there is, and that's a shame. Labeling data is so important and such an improvement in the SAV file format (over, say, SQL or CSV).
I guess the other way to deal with this is to not use codes, in favor of response strings, in the dataset. So, the Sex variable might have values: 'male', 'female', 'non-binary', etc. And I guess if you had your labels in a spreadsheet you could probably arrange to use INDEX/MATCH to replace the codes with response strings that would be clear to anyone looking at the data. Of course, that solves the labeling in a way, but when you import your data into PSPP, you then have to write a bunch of syntax to change those strings (of numeric variables like Likert responses) into numeric values to be used in analysis. And, I guess, ideally you'd want those numeric variables to have sensible value labels.
-Alan
On 1/21/2022 11:50 AM, jhwhite@techwriteinc.com wrote:
If I understand the issue correctly, variable labels are not being installed when importing some Excel files into PSPP. Is this correct?
From: Pspp-users <pspp-users-bounces+jhwhite=techwriteinc.com@gnu.org> On Behalf Of Alan Mead
Sent: Thursday, January 20, 2022 9:23 PM
To: Marek Ludwig <marek.ludwig@fh-potsdam.de>; pspp-users@gnu.org
Cc: Katja Behrndt <katja.behrndt@fh-potsdam.de>
Subject: Re: Import Codebook
I find applying labels to be very time-consuming, so maybe that's bad news for you. Maybe someone else will have a great idea.
But to make it as quick as possible, I'd recommend that you generate syntax and execute that syntax. I think that will be MUCH quicker than individually clicking and editing these values using the graphical user interface.
A lot of people are scared of syntax, but it's not so hard. An added advantage of doing it this way is that you easily fix an error by fixing the syntax and re-running it.
Also, if you have the information in a spreadsheet, I would try to generate the syntax using formulas in the spreadsheet. If column A contained the spss variable name (maybe "V1") and column B contained the variable label, then into cell C1 I would insert:
="variable labels "&A1&" '"&B1&"'."
(Note that there are single quotes, inside the double quotes, around B1 because it's a string.)
If A1 = V1 and B1 = Beschriftung then this would generate:
variable labels V1 'Beschriftung'.
And if you paste that into a syntax window, add the line "Execute." and run it, it would label this variable. You could paste 200 rows of Column C, add "Execute." and create the 200 variable labels very easily.
The value labels could be done similarly but I'd have to see the spreadsheet to devise the correct formula(s)...
This page describes the syntax:
This includes my solution and suggests an alternative (that may not work with PSPP):
https://www.reddit.com/r/spss/comments/mobw0z/import_excel_file_while_maintaining_variable/
Here are the relevant PSPP manual pages:
https://www.gnu.org/software/pspp/manual/html_node/VALUE-LABELS.html
https://www.gnu.org/software/pspp/manual/html_node/VARIABLE-LABELS.html
https://www.gnu.org/software/pspp/manual/html_node/MISSING-VALUES.html
-Alan
On 1/19/2022 9:01 AM, Marek Ludwig wrote:
Dear All,
we have read in a CSV dataset that we had generated from an Excel file. Unfortunately, the codebook got lost in the process, so that the columns for labels("Beschriftung"), value labels ("Wertelabels") and missing values ("Fehlende Werte") are empty. Since our dataset has over 200 variables, filling them in manually would be very time consuming. Is there an efficient, faster solution to read in the codebook or fill in these columns?
I would be very grateful for a hint!
Thanks a lot,
Marek
--Alan D. Mead, Ph.D.President, Talent Algorithms Inc.science + technology = better workershttps://talalg.comHofstadter's Law: It always takes longer than you expect, even when youtake into account Hofstadter's Law.--Alan D. Mead, Ph.D.President, Talent Algorithms Inc.science + technology = better workershttps://talalg.comGoing was easy. Keep on going was hard.-- Ursula K. Le Guin--Alan D. Mead, Ph.D.President, Talent Algorithms Inc.science + technology = better workershttps://talalg.comGoing was easy. Keep on going was hard.-- Ursula K. Le Guin
