export spatial weight matrix

[Souryabrata Mohapatra]

Hi,

I'm exploiting district-level unbalanced panel data.
I used the following command to generate a weight matrix w

spwmatrix gecon _CY _CX , wname(w) wtype(inv) cart dband(0 100) rowstand

And afterward, it said-

Inverse distance (alpha = 1) spatial weights matrix (13892 x 13892) calculated successfully and the following action(s) taken:

 - Spatial weights matrix  created as Stata object(s): w.

 - Spatial weights matrix has been row-standardized.

Now I want to use

xtmoran RYield, wname(w)

where w needs to be in dta format.

I don't know how to export the above-generated weight matrix as dta file.
I can't find the appropriate code.

Can you please advise,

Thank you,

[Andrew Musau]

help spmatrix export

[Souryabrata Mohapatra]

Thank you Andrew Musau,
But this export is in text format and Stata 17 crashes as I import it.
I need weights in a dta format to perform xtmoran.
Please advise.

[Andrew Musau]

I do not use these commands, but as long as you have Stata SE or Stata MP, it should cope with the size of your matrix.

spatial weights matrix (13892 x 13892)

I would suggest exporting as a CSV file, importing back to Stata and then splitting and destringing the values.

spwmatrix gecon _CY _CX , wname(w) wtype(inv) cart dband(0 100) rowstand
spmatrix export w using myfile.csv, replace
import delimited "myfile.csv", encoding(Big5) clear

[Souryabrata Mohapatra]

Thanks again Andrew Musau,

Which command would you use instead of spwmatrix?

I need to generate a weight matrix of the panel.
Then I need to do moran I and LM test to choose between lag and error models.
And I finally need to report the results of the suitable spatial model and show impact on a map.

Is it compulsory to have a balanced panel?

Awaits your reply,

[Andrew Musau]

As I said, I do not use these commands. The much help that I can give is to answer your question on how to convert the weighting matrix to a dataset. I would suggest that you start a new thread if you have a specific question on the commands or your model/data.

[Souryabrata Mohapatra]

Thank you Andrew Musau,

I created a 641 x 641 queen contiguity matrix in GeoDa and used the following command-

spwmatrix import using queen_weight.gal, wname(qweight) xport(qweight, txt)
insheet using weights.txt, delim(" ") clear

But for some reason, the order of the observations change as I convert it to txt and then export-import to dta import it to Stata.
Observation 431 in gal file is becoming observation 1 in Stata. Likewise, observation 619 in gal file is becoming observation 2 in Stata.
I wanted the observations in dta file to remain in the same order as gal file to do panel analysis later.

Can you please advise?

[Andrew Musau]

Once you import the matrix to Stata, you can sort using a variable that maps the original order to the order in the matrix. I do not know if the matrix rows are sorted alphabetically, you need to figure this out.

sysuse auto, clear
keep in 1/20
set seed 04042022
gen rsort= rnormal()
sort rsort
list make price mpg, sep(0)
*GENERATE ORIGINAL ORDER VARIABLE
gen order_old=_n

*GENERATE NEW ORDER VARIABLE (ALPHABETICAL SORTING)
sort make
gen order_new=_n
keep order*
tempfile order
save `order'

*NEW DATASET WITH ALPHABETICAL SORTING
sysuse auto, clear
keep in 1/20
sort make
list make price mpg, sep(0)
gen order_new=_n
*MERGE WITH DATASET WITH ORDER VARIABLES
merge 1:1 order_new using `order', nogen
sort order_old
list make price mpg, sep(0)

Res.:

. list make price mpg, sep(0)

     +----------------------------------+
     | make                 price   mpg |
     |----------------------------------|
  1. | Buick LeSabre        5,788    18 |
  2. | AMC Concord          4,099    22 |
  3. | Chev. Monza          3,667    24 |
  4. | Buick Regal          5,189    20 |
  5. | Chev. Nova           3,955    19 |
  6. | Cad. Eldorado       14,500    14 |
  7. | AMC Pacer            4,749    17 |
  8. | Cad. Seville        15,906    21 |
  9. | AMC Spirit           3,799    22 |
 10. | Chev. Impala         5,705    16 |
 11. | Dodge Colt           3,984    30 |
 12. | Buick Skylark        4,082    19 |
 13. | Buick Century        4,816    20 |
 14. | Chev. Chevette       3,299    29 |
 15. | Buick Electra        7,827    15 |
 16. | Chev. Malibu         4,504    22 |
 17. | Chev. Monte Carlo    5,104    22 |
 18. | Cad. Deville        11,385    14 |
 19. | Buick Opel           4,453    26 |
 20. | Buick Riviera       10,372    16 |
     +----------------------------------+

.

.
.
.
. *NEW DATASET WITH ALPHABETICAL SORTING

.
. sysuse auto, clear
(1978 Automobile Data)

.
. keep in 1/20
(54 observations deleted)

.
. sort make

.
. list make price mpg, sep(0)

     +----------------------------------+
     | make                 price   mpg |
     |----------------------------------|
  1. | AMC Concord          4,099    22 |
  2. | AMC Pacer            4,749    17 |
  3. | AMC Spirit           3,799    22 |
  4. | Buick Century        4,816    20 |
  5. | Buick Electra        7,827    15 |
  6. | Buick LeSabre        5,788    18 |
  7. | Buick Opel           4,453    26 |
  8. | Buick Regal          5,189    20 |
  9. | Buick Riviera       10,372    16 |
 10. | Buick Skylark        4,082    19 |
 11. | Cad. Deville        11,385    14 |
 12. | Cad. Eldorado       14,500    14 |
 13. | Cad. Seville        15,906    21 |
 14. | Chev. Chevette       3,299    29 |
 15. | Chev. Impala         5,705    16 |
 16. | Chev. Malibu         4,504    22 |
 17. | Chev. Monte Carlo    5,104    22 |
 18. | Chev. Monza          3,667    24 |
 19. | Chev. Nova           3,955    19 |
 20. | Dodge Colt           3,984    30 |
     +----------------------------------+

.
. gen order_new=_n

.
. *MERGE WITH DATASET WITH ORDER VARIABLES

.
. merge 1:1 order_new using `order', nogen
(label origin already defined)

    Result                           # of obs.
    -----------------------------------------
    not matched                             0
    matched                                20  
    -----------------------------------------

.
. sort order_old

.
. list make price mpg, sep(0)

     +----------------------------------+
     | make                 price   mpg |
     |----------------------------------|
  1. | Buick LeSabre        5,788    18 |
  2. | AMC Concord          4,099    22 |
  3. | Chev. Monza          3,667    24 |
  4. | Buick Regal          5,189    20 |
  5. | Chev. Nova           3,955    19 |
  6. | Cad. Eldorado       14,500    14 |
  7. | AMC Pacer            4,749    17 |
  8. | Cad. Seville        15,906    21 |
  9. | AMC Spirit           3,799    22 |
 10. | Chev. Impala         5,705    16 |
 11. | Dodge Colt           3,984    30 |
 12. | Buick Skylark        4,082    19 |
 13. | Buick Century        4,816    20 |
 14. | Chev. Chevette       3,299    29 |
 15. | Buick Electra        7,827    15 |
 16. | Chev. Malibu         4,504    22 |
 17. | Chev. Monte Carlo    5,104    22 |
 18. | Cad. Deville        11,385    14 |
 19. | Buick Opel           4,453    26 |
 20. | Buick Riviera       10,372    16 |
     +----------------------------------+

.

[Souryabrata Mohapatra]

Thanks again,

Since I have panel data, the order was important during export-import.
Say I have n number of ids then I need n×n weight matrix.
A balanced panel is a must for spatial analysis.

That command spwmatrix was changing the order while importing from gal file.
So I thought of using spmat insteed.
And guess what, it worked!

spmat import qweight using queen_weight.gal, geoda
spmat export qweight using queen_weight.txt
insheet using queen_weight.txt, delimiter(" ") clear

Only need minor corrections afterward.
And then

save "queen_weight.dta"
spmat dta queen_weight m*, normalize(row)
spmat summarize queen_weight, links detail

Now I can try xsmle, let's see.

Referring:

• https://doi.org/10.1177%2F1536867X1301300202
• https://doi.org/10.1177%2F1536867X1701700109