{smcl}
{hline}
help for {hi:spellsplit}{right:(E. Leuven)}
{hline}

{title:Splits records of time-span data}

{p 8 15}{cmdab:spellsplit} [{it:clist}], {cmdab:spell(}{it:start} {it:end}{cmdab:)} [{cmdab:by(}{it:varlist}{cmdab:)}]

{p}where {it:clist} is [{cmdab:(}stat{cmdab:)}] {it:varlist} [ [{cmdab:(}stat{cmdab:)} ... ]

{p}and stat is either {cmdab:mean} or {cmdab:sum}

{title:Description}

{p}{cmd:nojoin} breaks overlapping spells in disjoint adjacent spells. 

{p}{it:clist} with statistics
instructs how to treat variables when spells overlap; to sum them or take an average. 

{p 8 0} spellsplit x1 x2, spell(date0 date1) by(id)

is equivalent to

{p 8 0} spellsplit (mean) x1 x2, spell(date0 date1) by(id)
 
{p}As an example, if two overlapping spells
refer to a job and the variable is hours worked one would like to sum (to get the total hours worker), 
whereas if the variable is the worker's
age one would like to take the average (which will just be the worker's age).

{p}{cmd:spellsplit} Does the following

Suppose you have this

. l
           d0        d1          x
  1.   1Jan99    6Jan99          1
  2.   2Jan99    7Jan99         .2
  3.   4Jan99    8Jan99         .5
  4.   4Jan99   10Jan99         .1
  5.  16Jan99   21Jan99         .7

. spellsplit x, spell(d0 d1)

. l
              d0          d1          x   _count  
  1.   01Jan1999   02Jan1999          1        1  
  2.   02Jan1999   04Jan1999         .6        2  
  3.   04Jan1999   06Jan1999        .45        4  
  4.   06Jan1999   07Jan1999   .2666667        3  
  5.   07Jan1999   08Jan1999         .3        2  
  6.   08Jan1999   10Jan1999         .1        1  
  7.   16Jan1999   21Jan1999         .7        1  

where _count is the number of intercepting spells and x is averaged.
  
{title:Author}

{p}E. Leuven, Department of Economics, University of Amsterdam. If you observe any problems {browse "mailto:e.leuven@uva.nl"}.