[Return to Construction guide / CAMSIS home]
5.1.1 Handling pseudo-diagonals in SPSS: detailsFirst, to reiterate, this sample SPSS syntax shows the commands required given the file preparation already conducted.
One of the easiest ways of excluding the relevant husband-wife combinations from the analysis in SPSS is to create one or more dummy variables indicating that a particular occupational combination is to be regarded as a pseudo-diagonal, and then use a temporary select or filter command on that variable.
Firstly, as also described with the preliminary CA models (section 3), we identify pseudo-diagonal combinations from successive models by reviewing the scale scores from each stage of the model. In the first instance, the most extreme scores and their likely pseudo-diagonal combinations are often easily identified by the default SPSS graphics which plot male and female estimated scores in two dimensions. To be more thorough, however, we should also examine the ranking and values of scores for the male and female dimensions; this is easily achieved by, for example, pasting the SPSS table of results into Excel and then sorting on the relevant columns for the block of data. In scanning these results, pseudo-diagonals will again show up as occupational units ranked close to either extreme, with corresponding opposite gender units which may be linked by some plausible economic or institutional factor, also showing up as towards the same extreme.
In addition, there are several possible substantive shortcuts which can be taken in identifying pseudo-diagonals. It is often convenient to declare a priori all exact occupational title diagonals as pseudo-diagonals, or else if an occupational schema is nested into 'major' and 'minor' groups, it may be sensible to declare all occupational combinations where both partners are in the same group as pseudo-diagonals. In another example, if an occupational base unit at the title-by-status level is being used, it may also be possible to identify all combinations between one occupation, and all other occupations of a given status, or all combinations between specific status groups, as pseudo-diagonals. Examples of this last procedure might include the female occupation "Employee : Company secretary", with which all husband occupations which are "Self Employed: anything" could be identified; or the treatment of all husband-wife combinations where one spouse is self-employed and the other has an employment status of family assistant, as pseudo-dagionals.
The accompanying SPSS syntax illustrates how SPSS dummy variable computations can be used to build up a list of (potential) pseudo-diagonal husband-wife occupational combinations, that are subsequently selected for the analysis of the remaining occupations only. Note that because we conduct the CA model on the original SPSS dataset where revised and unrevised, and autorecoded and original category, occupational unit values are linked case by case, it is possible to specify the pseudo-diagonals in terms of the original data occupational unit numeric values (eg {h/w}bst), despite the fact that the subsequent model is run on the autorecoded and revised occupational unit values (eg {h/w}bst4).
The generation of an adequate CAMSIS model with SPSS CA, therefore, is achieved by successively running CA estimations, checking the output patterns for evident pseudo-diagonals, then updating the range of named pseudo-diagonals and running the CA estimation again.
Last modified 14 February
2002
This
document is maintained by
Paul Lambert (paul.lambert@stirling.ac.uk)