Splitting spell episodes that overlap a year

[Christopher Kappeler]

Dear Statalist,

I am working with episode data of the following form, where duration stands for the time span between begepi and endepi and UE is an unemployment dummy.

persnr   spell   begepi       endepi       duration   UE
1        1       01jan1985    31dec1985    365        0
1        2       01jan1986    31aug1986    243        0
1        3       01sep1986    21jul1989    1055       1
1        4       22jul1989    31jul1989    10         1
1        5       01aug1989    31jul1990    365        0
2        1       01jan2006    31dec2006    365        0
2        2       01jan2007    30jun2007    181        0
2        3       01jul2007    30jun2009    731        1

My aim is, to split each spell that overlaps one year (duration>365) and create out of this spell, news spells that do not overlap one year. In this case I want to split spell 3 of person 1 and person 2, respectively. To give a short example, in the end I would create a dataset (more or less) looking like this:

persnr   spell   begepi       endepi       duration   UE
1        1       01jan1985    31dec1985    365        0
1        2       01jan1986    31aug1986    243        0
1        3       01sep1986    21jul1989    1055       1
1        .       01sep1986    31dec1986    .          .
1        .       01jan1987    31dec1987    .          .
1        .       01jan1988    31dec1988    .          .
1        .       01jan1989    21jul1989    .          .  
1        4       22jul1989    31jul1989    10         1
1        5       01aug1989    31jul1990    365        1
2        1       01jan2006    31dec2006    365        0
2        2       01jan2007    30jun2007    181        0
2        3       01jul2007    30jun2009    731        1
2        .       01jul2007    31dec2007    .          .
2        .       01jan2008    31dec2008    .          .
2        .       01jan2009    30jun2009    .          . 

Does anyone of you have an idea how to do so? My feeling is that stsplit might be helpful, but I haven't arrived.
Thanks in advance for your help and your effort,

Best wishes, Christopher