Hi,

in this case (the variable is really nominal) you should create dummy-variables for each of the 4 categories. Make sure to use one of them (either a contentwise obvious one, or the one with the highest number of cases) as reference-category, which means that the respective category is not used as IV in your model.

Without being familiar with the exact question, I could imagine that "mother came alone" is a well interpretable reference category.

Best,

Stephan

MARIAN

There are only 3 dF for the 4 dummy variables, so only require 3 dummy variables... or alternately three orthogonal contrast variables. If dummy approach selected, then you might wish to have the three reflect where Mother not alone (with 0s for all three signifying Mom alone). If like contrasts, then first contrast could be Mother alone vs. sum of 1/3* each of the others, I.e., Mother alone vs accompanied. Not familiar with the three types of accompaniment, but if Health Outcomes are the interest, then one might wish to select two other orthogonal contrasts by some degree of personal attachment to Mom, perhaps: Family Member (esp. Daughter or Husband), Neighbor, Least or No personal attachment. As you are with UNICEF, my prospective ordering is framed by my cultural experience, so may be quite off, but may be obtainded with a quick ethographic survey if not already well known. Further ssuming Health Outcomes, I would note that accompanied would likely be required, if Mother were incapable of coming alone....so, an assessment of status on entrance would be of interest to tease-out that effect. Hopefully, my suggestions are not too far off your interest -- they are predicated on a series of health outcomes studies with which I had some involvement.

ALVAH