Replacing missing values in a panel structure

[Tharcisio Leone]

Hello everyone,

my data set contains 8,232 students (IDaluno) in a panel structure with T=5 (wave).
My dependent variable is test scores (Zprofic_mat) and the data are NOT completely balanced. See below that for the students 13648, 13649 and 13651, I have all the 5 test scores. But for the 13650, the test scores in T=1 and T=3 are missing.

I will estimate a value-added model. For this reason, I create the variable ZMat_LF which should import the first test score in the panel.

generate ZMat_LF = Zprofic_mat if wave==1
bysort IDaluno (wave): replace ZMat_LF=ZMat_LF[1]

. list IDaluno wave Zprofic_mat ZMat_LF in 1/20

     +---------------------------------------+
     | IDaluno   wave   Zprofic~t    ZMat_LF |
     |---------------------------------------|
  1. |   13648      1    .0793902   .0793902 |
  2. |   13648      2   -.6800427   .0793902 |
  3. |   13648      3    -1.84665   .0793902 |
  4. |   13648      4   -2.035958   .0793902 |
  5. |   13648      5   -1.590294   .0793902 |
     |---------------------------------------|
  6. |   13649      1    -1.98832   -1.98832 |
  7. |   13649      2   -.2193679   -1.98832 |
  8. |   13649      3   -1.813179   -1.98832 |
  9. |   13649      4   -.8301591   -1.98832 |
 10. |   13649      5   -.7455558   -1.98832 |
     |---------------------------------------|
 11. |   13650      2    .1301447          . |
 12. |   13650      4    .8231199          . |
 13. |   13650      5    1.288692          . |
 14. |   13651      1    .3280044   .3280044 |
 15. |   13651      2    .4950224   .3280044 |
     |---------------------------------------|
 16. |   13651      3    .1835145   .3280044 |
 17. |   13651      4    -.291254   .3280044 |
 18. |   13651      5    .4207084   .3280044 |
 19. |   13652      1    .4693756   .4693756 |
 20. |   13652      2    .0952681   .4693756 |
     +---------------------------------------+

The command above works fine for 13648, 13649 and 13651, but not for the 13650. So, I need a code now that in case of missing by ZMat_LF, than it replaces the test scores from T=2. If still missing, then T=3 and so on.

Does anyone have any idea of how can I create that?
Any advice would be highly appreciated!
Thanks in advance.

[William Lisowski]

* Example generated by -dataex-. For more info, type help dataex
clear
input int IDaluno byte wave float Zprofic_mat
13648 1  .0793902
13648 2 -.6800427
13648 3  -1.84665
13648 4 -2.035958
13648 5 -1.590294
13649 1  -1.98832
13649 2 -.2193679
13649 3 -1.813179
13649 4 -.8301591
13649 5 -.7455558
13650 2  .1301447
13650 4  .8231199
13650 5  1.288692
13651 1  .3280044
13651 2  .4950224
13651 3  .1835145
13651 4  -.291254
13651 5  .4207084
13652 1  .4693756
13652 2  .0952681
end
bysort IDaluno (wave): generate ZMat_LF = Zprofic_mat[1]
list, sepby(IDaluno) abbreviate(16) noobs

. list, sepby(IDaluno) abbreviate(16) noobs

  +-----------------------------------------+
  | IDaluno   wave   Zprofic_mat    ZMat_LF |
  |-----------------------------------------|
  |   13648      1      .0793902   .0793902 |
  |   13648      2     -.6800427   .0793902 |
  |   13648      3      -1.84665   .0793902 |
  |   13648      4     -2.035958   .0793902 |
  |   13648      5     -1.590294   .0793902 |
  |-----------------------------------------|
  |   13649      1      -1.98832   -1.98832 |
  |   13649      2     -.2193679   -1.98832 |
  |   13649      3     -1.813179   -1.98832 |
  |   13649      4     -.8301591   -1.98832 |
  |   13649      5     -.7455558   -1.98832 |
  |-----------------------------------------|
  |   13650      2      .1301447   .1301447 |
  |   13650      4      .8231199   .1301447 |
  |   13650      5      1.288692   .1301447 |
  |-----------------------------------------|
  |   13651      1      .3280044   .3280044 |
  |   13651      2      .4950224   .3280044 |
  |   13651      3      .1835145   .3280044 |
  |   13651      4      -.291254   .3280044 |
  |   13651      5      .4207084   .3280044 |
  |-----------------------------------------|
  |   13652      1      .4693756   .4693756 |
  |   13652      2      .0952681   .4693756 |
  +-----------------------------------------+

Please note the use of the dataex command to present example data in a form that can be copied and pasted into a do-file.