# linking two variables in a household roster

Discussion in 'SPSS' started by Mattias, Nov 7, 2011.

1. ### MattiasGuest

Dear Group,

I posted a question last week regarding how to link mothers and their
children from a household roster we used in our survey. I realize now
that I wrote too early and am now trying to get help a second time.
This time I am hopefully being more specific and informative. We are
looking for ways of making SPSS do the following for us;

We are trying to link mothers with their children as we want to look
at the odds of a birth being a son as repeated measures/outcomes by
each mother. We first determine that household members are eligible
women, i.e. eligible to be respondents to our women’s schedule. They
need to have a value of 1 for Yes on our variable “eligible person”
and a value of 2 for Female on our variable “Sex”. Our roster included
the variable “relationship to head of household”. This is the variable
that links mothers to their children. (Yes, it does so in an
unnecessarily roundabout way, but since we missed including a simple
variable determining who is the mother of each child, this seems to be
the only variable that we can use to link them.) Each eligible woman
and each of her children have a value determining how they are related
to a third person in the household – the head of the household. For
example, if the eligible woman’s relationship to the head of the
household is “daughter-in-law” (i.e. that she is the daughter-in-law
of the head of the household), then her children will be coded as
“niece or nephew” for her or his relationship to the head of the
household. We can list out all possible combinations (if eligible
woman is the daughter of the head of the household, her children will
be the grandchildren of the head of the household, and so on) and thus
have one link between each eligible woman and her children.

We want to construct a variable for each birth by each mother named,
say, “kid_sex”. Since we are doing a repeated measures design, we
eventually want each child to be assigned to its mother as a birth
(1=Son, 0=Daughter) with each mother having one row for each child. We
want to also draw within-subject variables such as child’s age and
parity from the roster.

I guess in shorter form, the steps we need to ask the software to take
would be;

1. Select only eligible women based on them being (a) an eligible
person, and (b) a woman.
2. Tell SPSS to assign children to this eligible woman based on her
“relationship to head of household”, from which we know what
“relationship to head of household” her children will have, and thus
who her children are.
3. Construct one variable for each birth/child.

The question is; how can we make SPSS construct a variable for each
birth by each mother based on this information?

Thank you,
Mattias

Mattias, Nov 7, 2011

2. ### DavidGuest

Please provide an example of how the data appear, the variables useful
for creating this link and how the final data should appear. What
confuses me most is how you intend to have this sorted correctly with
such indirect and ambiguous structure.
What if there are two daughters with children? Sounds like all the
kids are grandkids. Who's the momma? Sounds like a rather sloppy
data collection methodology!

David, Nov 7, 2011

3. ### Rich UlrichGuest

The computer is not going to be able to do magic.

You have to list out the relations that you intend to use to
create codes, as elegantly as possible, and then you will
note any ambiguities that you can imagine. But you do have
to start by being correct with what you list. - The example
you state in the sentence just above is wrong; the "niece
or nephew" must be a grandchild since the intermediary role
of "daughter-in-law" is from the same generation as "daughter"
which is described in the next sentence.

It may be that "generation" is a logical category that you don't
correct assignments. - Doing a test on one variable instead of
several is part of what I had in mind when I used the word
"elegant."

Why don't you show some data, and some of what you construct,
and mention what goes wrong?

Rich Ulrich, Nov 8, 2011