birth weight low, to decide whether to treat it as a separate or multiple pregnancy.
The results are logged in the table attached to Appendix 7.
(v) A small number of outcome dates (all in BCS70) look as though they may have
been entered erroneously, i.e the year was before 1970 or when the CM was
extraordinarily young. See Appendix 8.
(vi) A small number of respondents answered ‘Yes’ to the question ‘Have you ever
been, or got anybody pregnant?”, but gave no information about any dates or indeed
whether the pregnancy went to full term. In these cases, there is no basis for
attempting any re-ordering. See Appendix 9.
Algorithm for correcting the errors
Step 1
Create a vector BIRDAT, containing 40 variables, corresponding to each of the forty
birth slots, using the SPSS ‘YRMODA’ function, which produces a value equal to the
number of days elapsed from a fixed time-point to the date in question. This allows
all outcome dates to be compared chronologically, except for cases mentioned in
paras (ii)-(vi) above, where there is an incomplete, missing or dubious date.
Step 2
In the case of incomplete dates where at least a valid year has been entered, assign a
value for BIRDATn by imputing ‘15’ for a missing day, and ‘6’ for a missing month.
Although this might result in two successive outcome dates appearing to be less than
nine months apart, it should, to all intents and purposes, be impossible for them to get
into the wrong order as a result of the imputation (for another outcome with a
specified date to happen in the same year, the other event would have to be in the first
three months or the last three months of the year, except in the rare case of a
spontaneous miscarriage of one child from a multiple pregnancy). It should be noted
that this imputation is only a temporary device for ordering the birth slot data
correctly. The values of the actual outcome date variables (prege-40, preged-40,
pregem-40, pregey-40) are not altered.
Step 3
In the case of ongoing pregnancies, assign a value for BIRDATn by imputing the
date 30th September 2000. As all interviews had been completed by this date, it
would be impossible for there to be a subsequent outcome, ensuring the ongoing
pregnancy would be placed as ‘most recent’ in the corrected ordering. There were
some ongoing pregnancies where two birthslots had been filled in, but one can assume
these were cases where it was already known the mother was expecting twins (see
Appendix 5). The imputation in these cases would result in two identical outcome
dates of 30/9/2000, which would correctly enable them to be flagged up as twins
under Step 7 below.