Centre for Longitudinal Studies

loop #i=1 to 8.

loop #j=1 to 4.

if ((birdat(5*#i-4+#j)-birdat(5*#i-4) ge 273) or (birdat(5*#i-4)-birdat(5*#i-4+#j) ge
273))bunched=1.

end loop.

*Flag up all cases where there is an ongoing pregnancy.

compute stilpreg=0.

format stilpreg (f1.0).

loop #i=1 to 40.

if (vprega(#i)=5)stilpreg=1.

end loop.

* Get all valid birth dates in chronological order, up to a maximum of 8 births.

* Use 'key' vector to keep track of which birthslot the date was originally placed in, so as to

* line up all the other pregnancy variables corresponding with this birth into their correct slots.

* After each of the 8 iterations resulting in determining a 'key' number, set the outcome date

* for the slot that key points to, to zero, so it will no longer be the most recent in the next
iteration.

loop #j=1 to 8.

compute outdat(#j)=0.

compute key(#j)=0.

loop #i=1 to 40.

Do if (birdat(#i)>outdat(#j)).

compute outdat(#j)=birdat(#i).

compute key(#j)=#i.

end if.

end loop.

if (key(#j) ne 0)birdat(key(#j))=0.

end loop.

* As a check that everything worked OK, look at whether any birdat values are still non-zero.
compute missdany=0.

format missdany (f1.0).

loop #i=1 to 40.

If (birdat(#i) ne 0)missdany=1.

end loop.

* Create variable truemult to flag up all cases with true multiple births: i.e. those where the
outcome date is the same,

* or no more than 3 days apart.

compute truemult=0.

format truemult (f1.0).

loop #i=1 to 7.

if ((outdat(#i)-outdat(#i+1) le 3) and (outdat(#i+1) ne 0))truemult=1.

end loop.

* Create warning flag variable oddmult to flag outcome dates less than nine months apart, but
greater than 3 days apart.

compute oddmult=0.