mlogit transition

[Guest]

Cheers, Sample data here:

clear
set obs 16
egen id = seq(), from(1) to(4) block(4)
egen group = seq(), from(0) to(1) block(8)
gen job = .
replace job = 1 in 1
replace job = 1 in 2
replace job = 1 in 3
replace job = 2 in 4
replace job = 2 in 5
replace job = 1 in 6
replace job = 1 in 7
replace job = 2 in 8
replace job = 1 in 9
replace job = 1 in 11
replace job = 3 in 12
replace job = 1 in 13
replace job = 2 in 14
replace job = 3 in 15
replace job = 3 in 16

Basically I want to estimate transition for each group between job categories where job equals to 1 mean not employed, job equals to 2 means interviewing, job equals to 3 means interviewing. So I want to estimate the probability of transition between all the job categories for each group (group 0 and group 1). I read to use mlogit command but I do not know how to make it specify the transition from a particular job group.

 
 bysort group: mlogit job

[Rich Goldstein]

I am not sure I understand; in particular, I don't understand why you are using -mlogit- for transition probabilities; here is a different way to look at it (which might be completely wrong if I have mis-understood you); after reading in the data (thank you for supplying), I added a time variable as follows:

bys id: gen t=_n

then I -xtset- the data and used -xttrans-:

xtset id t
bys group: xttrans job

be sure to look at the help file (or manual entry) for -xttrans- to see if it is what you want