suest singular variance matrix

[filippo dare]

Dear all,

I'am trying to run this code:

reg outcome i.post i.cell i.data_periods i.post#i.data_periods treated if split_var==1,
estimates store A

reg outcome i.post i.cell i.data_periods i.post#i.data_periods treated if split_var==0,
estimates store B
suest A B, cluster(double_cluster)
test [A_mean]treated=[B_mean]treated

Sorry if don't usa dataex but the code takes half an hour to work. At the suest point it reminds me the error "Variance matrix highly asymmetric or non-singular". I read a lot of previous posts on this issue but no one seems my problem. I attached with dataex my data.

* Example generated by -dataex-. For more info, type help dataex
clear
input byte outcome float(post data_periods treated cell individual_id cluster_var split_var)
4 0 1 0  1 311  770 1
4 0 1 0  2 405  995 1
1 0 1 0  3  28   73 1
6 0 1 0  3 478 1170 0
6 0 1 0  4 434 1064 1
2 0 1 0  5  92  227 0
2 0 1 0  6 365  900 0
1 0 1 0  6  63  160 1
5 0 1 0  6 117  287 1
4 0 1 0  7 125  311 0
3 1 1 1  8 242  605 1
1 0 1 0  8 175  417 1
4 0 1 0  9 512 1249 1
1 0 1 0 10 158  375 0
9 0 1 0 11 352  868 1
0 0 1 0 11 106  258 0
1 0 1 0 12 297  730 0
3 0 1 0 12 121  300 0
0 0 1 0 13 237  591 1
2 0 1 0 14 109  266 0
1 0 1 0 14 215  533 1
0 0 1 0 15 364  897 0
3 1 1 1 16 254  627 1
1 0 1 0 17 130  323 1
0 0 1 0 17 207  507 1
5 1 1 1 18 417 1022 0
1 0 1 0 19  34   88 0
1 0 1 0 20 456 1117 1
1 0 1 0 21 421 1033 0
4 0 1 0 22  47  121 0
0 1 1 1 23 238  594 1
0 1 1 1 24 319  789 0
1 0 1 0 25 429 1052 1
1 1 1 1 26 261  645 1
5 0 1 0 26 443 1085 1
3 0 1 0 27  16   39 1
2 0 1 0 28 218  542 0
2 0 1 0 29 146  356 0
0 0 1 0 30  60  155 0
1 0 1 0 31 373  918 1
1 0 1 0 32  82  200 0
0 0 1 0 33  96  235 0
2 0 1 0 33 455 1116 0
2 0 1 0 33 252  625 1
0 0 1 0 34 282  695 0
0 0 1 0 35 153  366 1
1 0 1 0 36  99  242 1
0 0 1 0 36 241  602 1
1 0 1 0 36 317  784 0
2 0 1 0 36 262  647 1
1 0 1 0 37 320  791 0
4 0 1 0 37 186  443 1
4 0 1 0 37 280  690 1
1 0 1 0 37 151  363 1
1 1 1 1 37 303  751 1
0 0 1 0 37 189  448 1
1 1 1 1 37 314  778 1
2 0 1 0 38 401  988 0
0 0 1 0 39  10   21 1
4 0 1 0 39 170  407 1
0 1 1 1 40  94  232 1
1 0 1 0 41 520 1270 0
2 1 1 1 41  63  161 1
1 1 1 1 42 396  976 0
1 0 1 0 43 334  823 0
0 0 1 0 43 349  861 0
0 0 1 0 44  69  176 1
3 0 1 0 44 103  251 1
0 1 1 1 45 152  364 1
0 1 1 1 45 179  428 0
0 0 1 0 45 441 1083 1
5 0 1 0 46 492 1209 1
2 1 1 1 46 122  303 0
6 0 1 0 47 335  826 0
0 0 1 0 47  39  100 1
5 0 1 0 47 102  248 0
3 0 1 0 47 462 1131 0
1 0 1 0 47  95  234 1
1 0 1 0 48  48  124 0
0 0 1 0 49 135  335 0
0 0 1 0 50 493 1210 0
7 1 1 1 50  76  189 0
0 1 1 1 51 158  376 0
1 0 1 0 52 161  384 0
2 0 1 0 53 494 1211 1
0 0 1 0 54 267  657 0
2 0 1 0 55 362  895 1
1 0 1 0 55 466 1140 1
0 0 1 0 55 497 1214 1
0 0 1 0 55 388  958 0
0 0 1 0 55 131  326 0
0 0 1 0 55 250  622 1
0 1 1 1 56 424 1040 1
6 1 1 0 57 311  771 1
3 0 1 0 58 150  362 1
3 1 1 1 59   5   11 1
0 1 1 1 60  34   89 1
8 0 1 0 60 511 1248 1
1 0 1 0 61 412 1011 1
7 1 1 1 61 405  996 0
end
label values outcome eurod
label def eurod 0 "Not depressed", modify

[Joseph Coveney]

What does

regress outcome i.split_var##i.cell i.split_var##i.post##i.data_periods i.split_var##i.treated

or

hetregress outcome i.split_var##i.cell i.split_var##i.post##i.data_periods i.split_var##i.treated, ///
    het(i.split_var)

give you?

suest A B, cluster(double_cluster)

double_cluster isn't in your data snippet, not that it matters much, given how uninformative the snippet is.