Changing matrix entries to 1

[Pawel Charasz]

Hello,

I am trying to edit a spatial weighting matrix created by the spmatrix create command in Stata. I would like to replace all the non-zero entries (>0) with 1, but this seems to not be working.

I have the following code to do so, where spatial_weights is a weighting matrix created with spmatrix create:

spmatrix matafromsp W id = spatial_weights

mata:
    for (i=1; i<=rows(W); i++) {
        for (j=1; j<=cols(W); j++) {
            if (W[i,j]:>0) W[i,j]=1
        }
    }
end

spmatrix spfrommata spatial_weights = W id, replace

What I get as a result is a matrix of indeed same entries for all the previously non-zero entries, but they all take the value of .01065647570978. Any ideas of what am I doing wrong?

[Joseph Coveney]

Maybe try something like the following?

.ÿclearÿ*

.ÿlocalÿline_sizeÿ`c(linesize)'

.ÿsetÿlinesizeÿ72

.ÿmatrixÿinputÿWÿ=ÿ(0ÿ0.1ÿ0ÿ\ÿ0ÿ0.2ÿ0.1)

.ÿmatrixÿlistÿW

W[2,3]
ÿÿÿÿc1ÿÿc2ÿÿc3
r1ÿÿÿ0ÿÿ.1ÿÿÿ0
r2ÿÿÿ0ÿÿ.2ÿÿ.1

.ÿmata:
-------------------------------------------------ÿmataÿ(typeÿendÿtoÿexit
>ÿ)ÿ--------------------------------------------------------------------
:ÿWÿ=ÿst_matrix("W")

:ÿW
ÿÿÿÿÿÿÿÿ1ÿÿÿÿ2ÿÿÿÿ3
ÿÿÿÿ+----------------+
ÿÿ1ÿ|ÿÿÿ0ÿÿÿ.1ÿÿÿÿ0ÿÿ|
ÿÿ2ÿ|ÿÿÿ0ÿÿÿ.2ÿÿÿ.1ÿÿ|
ÿÿÿÿ+----------------+

:ÿWÿ=ÿW:>0

:ÿW
ÿÿÿÿÿÿÿ1ÿÿÿ2ÿÿÿ3
ÿÿÿÿ+-------------+
ÿÿ1ÿ|ÿÿ0ÿÿÿ1ÿÿÿ0ÿÿ|
ÿÿ2ÿ|ÿÿ0ÿÿÿ1ÿÿÿ1ÿÿ|
ÿÿÿÿ+-------------+

:ÿst_matrix("W",ÿW)

:ÿend
------------------------------------------------------------------------

.ÿmatrixÿlistÿW

W[2,3]
ÿÿÿÿc1ÿÿc2ÿÿc3
r1ÿÿÿ0ÿÿÿ1ÿÿÿ0
r2ÿÿÿ0ÿÿÿ1ÿÿÿ1

.ÿsetÿlinesizeÿ`line_size'

.

[Sebastian Kripfganz]

Be default, the spmatrix command applies a spectral normalization of the matrix. Use the option normalize(none) if you do not want any normalization (which would be unusual) or normalize(row) if you want a row standardization (which might have undesired effects on the implied network structure and will also change your entries to different numbers than one).

[Pawel Charasz]

Be default, the spmatrix command applies a spectral normalization of the matrix. Use the option normalize(none) if you do not want any normalization (which would be unusual) or normalize(row) if you want a row standardization (which might have undesired effects on the implied network structure and will also change your entries to different numbers than one).

Sebastian Kripfganz , thank you, you are spot on. I thought for some reason that normalization is only applied with spmatrix create and not spmatrix spfrommata, and that I need to manually normalize it with spmatrix normalize after my transformation.That's why I didn't expect my matrix to already be normalized.