Diagonal

[Marco Siciliano]

I am modelling a huge database and I have to create a new variable that contains the value of the diagonal of the current database.
Basically on the x-axis and y-axis I have the same variables and I want to select the interaction term - you can think about it as the variance in the variance-covariance matrix. I have tried with MATA and the function diagonal() but I get errors.
Any help with the code? I would really appreciate it!

[Leonardo Guizzetti]

You're question is unclear. If you want the diagonal of a matrix, the vecdiag() and diag() functions can be useful. Here's an example of those functions:

mat X = (1,2,3 \ 4,5,6 \ 7,8,9)
mat list X
mat dv = vecdiag(X)
mat list dv
mat D = diag(vecdiag(X))
mat list D

Result

. mat X = (1,2,3 \ 4,5,6 \ 7,8,9)

. mat list X

X[3,3]
    c1  c2  c3
r1   1   2   3
r2   4   5   6
r3   7   8   9

. mat dv = vecdiag(X)

. mat list dv

dv[1,3]
    c1  c2  c3
r1   1   5   9

. mat D = diag(vecdiag(X))

. mat list D

symmetric D[3,3]
    c1  c2  c3
c1   1
c2   0   5
c3   0   0   9

If this does not accomplish what you want, then please spend some time reading the FAQ, in particular section 12 on how to ask effective questions. You are encouraged to post a reproducible data example using -dataex- which is included with fully updated versions of Stata 14.2, 15.1, 16 and 17.

[Maarten Buis]

Typically, when using such "tabular" data in a program like Stata, you will want that data in long or vectorized form, e.g.:

. use "http://fmwww.bc.edu/repec/bocode/m/mob.dta", clear
(mobility table from the USA collected in 1973)

. list, sepby(row)

     +------------------------------------------+
     |             row               col    pop |
     |------------------------------------------|
  1. | upper nonmanual   upper nonmanual   1414 |
  2. | upper nonmanual   lower nonmanual    521 |
  3. | upper nonmanual      upper manual    302 |
  4. | upper nonmanual      lower manual    643 |
  5. | upper nonmanual              farm     40 |
     |------------------------------------------|
  6. | lower nonmanual   upper nonmanual    724 |
  7. | lower nonmanual   lower nonmanual    524 |
  8. | lower nonmanual      upper manual    254 |
  9. | lower nonmanual      lower manual    703 |
 10. | lower nonmanual              farm     48 |
     |------------------------------------------|
 11. |    upper manual   upper nonmanual    798 |
 12. |    upper manual   lower nonmanual    648 |
 13. |    upper manual      upper manual    856 |
 14. |    upper manual      lower manual   1676 |
 15. |    upper manual              farm    108 |
     |------------------------------------------|
 16. |    lower manual   upper nonmanual    756 |
 17. |    lower manual   lower nonmanual    914 |
 18. |    lower manual      upper manual    771 |
 19. |    lower manual      lower manual   3325 |
 20. |    lower manual              farm    237 |
     |------------------------------------------|
 21. |            farm   upper nonmanual    409 |
 22. |            farm   lower nonmanual    357 |
 23. |            farm      upper manual    441 |
 24. |            farm      lower manual   1611 |
 25. |            farm              farm   1832 |
     +------------------------------------------+

In my example each observation is a cell in the table and the variables indicating what row and column a cell is on are called row and col. The content of the cell is in my example in the variable pop. A cell is on the main diagonal (from top left to bottom right) when row == col. You can make a variable indicating when that is the case as follows: gen diag = (row==col) . This will make a variable called diag, which takes the value 1 when it is on the diagonal an 0 otherwise. There are several tips and tricks on dealing with this type of data in this talk I gave at the 2015 Stata Users' Group meeting in London: http://maartenbuis.nl/presentations/london15b.pdf