Question: how to compute a variable value based on its last value in a panel dataset

[Di Yang]

Dear All,

I am trying to do the calculation illustrated in Excel with Stata. Specifically, I would like to compute values in column M. But as you could see in the formula line, each value in column M needs the previous value as input to compute. For example, the value of M6 depends on the value of M5, and so on. I am struggling to achieve this using Stata with the codes below. It always give me missing values. I really appreciate it if someone could help me with this.


Many thanks in advance for your assistance,
Di

     *generate counterfactuals (M column) 
    gen cf1 = 0 
    replace cf1 = a0 + a1*cf1'[_n-1] + b2*shock2 + b3* shock3 + et

[Carlo Lazzaro]

Di:
I do hope that the follows toy-example helps:

. set obs 10
number of observations (_N) was 0, now 10

. g A=runiform()

. g B=runiform()

. g C=runiform()

. g D= A+ (B* C[_n-1])

. list

     +-------------------------------------------+
     |        A          B          C          D |
     |-------------------------------------------|
  1. | .3488717   .2047095   .9319346          . |
  2. | .2668857   .8927587   .4548882   1.098878 |
  3. | .1366463   .5844658   .0674011   .4025129 |
  4. | .0285569   .3697791   .3379889   .0534804 |
  5. | .8689333   .8506309   .9748848   1.156437 |
     |-------------------------------------------|
  6. | .3508549   .3913819   .7264384   .7324072 |
  7. | .0711051   .1196613   .0454151   .1580317 |
  8. |  .323368   .7542434   .7459667    .357622 |
  9. | .5551032   .6950234   .4961259   1.073567 |
 10. |  .875991   .6866152   .7167162   1.216639 |
     +-------------------------------------------+

.

[Di Yang]

Hi Carlo,

Many thanks for your reply.

Do you know what I shall do if I want to compute D= A+ (B* D[_n-1]) with D = 0 for the first observation? The key difficulty I am facing is that I need to generate a new variable and the value of this variable in time t depends not only on other variables at given time t but its previous value in t-1 as well. I don't know how to incorporate the latter part in the calculation.

Thank you,
Di

[Carlo Lazzaro]

Di:
you can't.
See the following toy-example that hopefully mimicks what you're after:

. set obs 1
number of observations (_N) was 0, now 1

. g D=0

. g A=1

. g B=A+D[_n-1]
(1 missing value generated)

. list

     +-----------+
     | D   A   B |
     |-----------|
  1. | 0   1   . |
     +-----------+

.