How to Begin?



Hi All:

I'm working with a large national data set and want to make sure that I
start out on the right track. So, I'd appreciate feedback on which
analyses would be most useful (I'm primarily using SPSS, but also Excel
for charting and basic descriptive statistics, as requested by client).
One of the issues is that everything is significant with such a large
sample size, so tests of significance, in and of themselves, are not
particularly useful.

My main variables of interest are related to work type:
FT/PT - categorical, 2 levels
Permanency - categorical, 4 levels
Combined (i.e. FT Perm, FT Temp, etc.)- categorical, 4 levels

Basically, I'm interested in the relationship of these variables to
other variables regarding employment and employee characteristics and
interactions; for example: wages, tenure, gender, age. There are about
15 of these variables, most of which either are discrete or I've
transformed into discrete variables, although a few, less central ones
remain continuous.

Additionally, I'm looking at changes in these relationships over time,
on a yearly basis. This is a bit of an issue with respect to the layout
and collection of the data, which is from a national Labour Force
Survey. Participants provided information each month in relation to one
reference week that month. Thus, each case in the data set refers to
one participant's responses in terms of one week each month. This isn't
as much of a problem when working with means for the year, but it is an
issue when using statistics that refer to yearly frequencies. Further,
each household participates for only 6 months and then is replaced by a
commensurate household, so this isn't a within-subjects repeated
measures design.

My clients aren't looking for advanced or overly technical results, and
would be quite satisfied if I just reported basic descriptive stats.
However, with the number of variables and potential interactions, I'd
like to know which ones to highlight and which of the differences and
trends in descriptive stats are valid phenomena. I've been looking into
multiway frequency analysis or logistic regression, but I'm not sure
the data is set up appropriately, as is, and I'm also a bit confused by
the different SPSS options and output.

I'd appreciate any suggestions. Thanks in advance, Joy.

.