I have been trying to learn the Codebook methodology by Yair Alan Griver,
and I find it very tastefull. Still I feel like an amateur so I need some hints
now and then.
Right now I am trying to build an application to store queries, precoded answers, surveys
(consisting of queries from that table) and the answers to the surveys.
The structure should be in a way so that I can put it on the web.
Queries belong to everyone or to a specific custom acount.
There will also be a way to track respondents and to follow up response.
The creation of the query database is straightforward, like customer acounts.
Problems arise when I try to make a structure out of the answers. One query can generate more
than one reply (multiple response or multipunch) and I want to be able to have full control
over the variable names that will be passed to a statistical package for further analysis.
One program that can handle part of this is In2itive from Denmark (now a part of SPSS).
However, I don't like the way they set the variables names.
Example: Mark the brands you know about.
Volvo
SAAB
Mercedes
Audi
Dodge
Chrysler

I want to be able to give all of the diferent alternatives their own variable name. They can
consist of eight characters. In this example lets say I give them the names:
volvo
saab
mercedes
audi
dodge
chrysler

They are all coded 0 for no response and 1 for respons.
To build a transactional model for replies I would have one record for each reply and none for
none replies.
Since a statistical program like SPSS works with a flat file (matrix) I have to create records
even for non replies.
In2itive seems to give the variable names from the query wich always have a name and wich is
easy to implement in a program.
Let's say that we call this question Carbrand. We could then call alternative 1 (Volvo) Carbr#a,
alternative 2 (SAAB) Carbr#b and so on.
If I do this survey once a month and I have to update the list of brands, and if I have to change
the order of the alternatives, I am facing problems. Carbr#a becomes Carbr#b and so on.
Therefore I want to put a little extra effort in solving this.
When I use SPSS I always have to set up my variable scheme and all the labels every time I read an
ASCII-file or a DBF-file or whatever file. Since In2itive generates this setup with the names
given in the database I don't seem to have control over the names in a multipunch query since it
just trunkates the name of the query, add a # and increments or indexes the variables.
Since I expect the query database to be large I can't put the variablenames here, since this will
limit people in naming their variables since the names have to be unique.
The most natuaral place to set them is when you have built a survey out of some queries.
If someone can follow my thoughts and have any ideas, please tip me!
I would be very greatfull if someone can give me examples of databases like this.
I have seen In2itive, Quantime and DOBS, and I am impressed but not amused.