My Input: a type 2 association (RC model) with husband’s education and wife’s education:
ass2, with model 5a. Note that model 5e (which is supposed to yield unconstrained scores for row and col) yields exactly the same answer. I don’t know why.
Input:
dim 4 4
lab H W
mod
dat [32016 33374 8407 988
28370 137876 43783 8446
7051 48766 61633 18195
984 13794 28635 51224]
Output:
LEM: log-linear and event history analysis with missing data.
Developed by Jeroen Vermunt (c),
Version 1.0 (
*** INPUT ***
man 2
dim 4 4
lab H W
mod
dat [32016 33374 8407 988
28370 137876 43783 8446
7051 48766 61633 18195
984 13794 28635 51224]
*** STATISTICS ***
Number of iterations = 62
Converge criterion = 0.0000009227
X-squared = 11374.7967 (0.0000)
L-squared = 10528.7995 (0.0000)
Cressie-Read = 11029.4300 (0.0000)
Dissimilarity index = 0.0622
Degrees of freedom = 4
Log-likelihood = -1235422.80534
Number of parameters = 11 (+1)
Sample size = 523542.0
BIC(L-squared) = 10476.1260
AIC(L-squared) = 10520.7995
BIC(log-likelihood) = 2470990.4628
AIC(log-likelihood) = 2470867.6107
WARNING: no information is provided on identification of parameters
*** FREQUENCIES ***
H W observed estimated std. res.
1 1 32016.000 29323.716 15.722
1 2 33374.000 38493.237 -26.092
1 3 8407.000 6653.414 21.498
1 4 988.000 314.697 37.955
2 1 28370.000 33184.593 -26.430
2 2 137876.000 125988.448 33.491
2 3 43783.000 51752.316 -35.031
2 4 8446.000 7549.708 10.315
3 1 7051.000 5547.488 20.186
3 2 48766.000 56545.669 -32.716
3 3 61633.000 51950.996 42.478
3 4 18195.000 21600.815 -23.173
4 1 984.000 365.203 32.380
4 2 13794.000 12782.646 8.945
4 3 28635.000 32101.274 -19.346
4 4 51224.000 49387.781 8.263
*** LOG-LINEAR PARAMETERS ***
* TABLE HW [or P(HW)] *
effect beta exp(beta)
main 9.6424 1.54E+0004
H
1 -0.7927 0.4526
2 0.8419 2.3207
3 0.4582 1.5812
4 -0.5073 0.6021
W
1 -0.8381 0.4326
2 1.0328 2.8088
3 0.5805 1.7870
4 -0.7752 0.4606
type 2 association (row=H column=W)
association 5.1277
row -0.6590 -0.2212 0.1859 0.6944
adj row -1.4923 -0.5010 0.4209 1.5724
column -0.6731 -0.2000 0.1856 0.6874
adj column -1.5241 -0.4528 0.4204 1.5566
*** (CONDITIONAL) PROBABILITIES ***
* P(HW) *
1 1 0.0560
1 2 0.0735
1 3 0.0127
1 4 0.0006
2 1 0.0634
2 2 0.2406
2 3 0.0989
2 4 0.0144
3 1 0.0106
3 2 0.1080
3 3 0.0992
3 4 0.0413
4 1 0.0007
4 2 0.0244
4 3 0.0613
4
4 0.0943
-----------------------------------------------------------------------
*And now, something different, using LEM to estimate the uniform
association, or linear by linear association model. Notice that the residual df and the LRT chisquare are exactly as I calculated them
from STATA, but the association parameter here is listed differently, so it is
scaled differently in LEM.
ass2, with model type 2a.
LEM:
log-linear and event history analysis with missing data.
Developed by
Jeroen Vermunt (c),
Version 1.0 (
*** INPUT ***
man 2
dim 4 4
lab H W
mod
dat [32016 33374
8407 988
28370 137876 43783 8446
7051 48766 61633 18195
984 13794 28635 51224]
*** STATISTICS
***
Number of iterations = 38
Converge criterion = 0.0000006657
X-squared = 13083.9241 (0.0000)
L-squared = 12561.3499 (0.0000)
Cressie-Read = 12863.9121 (0.0000)
Dissimilarity index = 0.0649
Degrees of freedom = 8
Log-likelihood = -1236439.08052
Number of parameters = 7 (+1)
Sample size = 523542.0
BIC(L-squared) = 12456.0029
AIC(L-squared) = 12545.3499
BIC(log-likelihood) = 2472970.3396
AIC(log-likelihood) = 2472892.1610
WARNING: no
information is provided on identification of parameters
*** FREQUENCIES
***
H W
observed
estimated std. res.
1 1 32016.000 28854.715 18.610
1 2 33374.000 40075.847 -33.478
1 3
8407.000 5506.474 39.088
1 4
988.000 347.959 34.312
2 1 28370.000 33991.074 -30.489
2 2 137876.000 128324.718 26.663
2 3 43783.000 47927.051 -18.929
2 4
8446.000 8232.188 2.357
3 1
7051.000 5110.281 27.148
3 2 48766.000 52440.848 -16.047
3 3 61633.000 53237.739 36.385
3 4 18195.000 24856.140 -42.250
4 1
984.000 464.931 24.073
4 2 13794.000 12968.588 7.248
4 3 28635.000 35786.736 -37.805
4 4 51224.000 45416.712 27.250
*** LOG-LINEAR
PARAMETERS ***
* TABLE HW [or P(HW)] *
effect beta exp(beta)
main 9.6597 1.56E+0004
H
1
-0.8261 0.4378
2
0.8377 2.3110
3
0.4428 1.5570
4
-0.4544 0.6348
W
1
-0.8135 0.4433
2
1.0150 2.7593
3
0.5301 1.6991
4
-0.7316 0.4812
type 2 association
(row=H column=W)
association 4.9998
*** (CONDITIONAL)
PROBABILITIES ***
* P(HW) *
1 1
0.0551
1 2
0.0765
1 3
0.0105
1 4
0.0007
2 1
0.0649
2 2
0.2451
2 3
0.0915
2 4
0.0157
3 1
0.0098
3 2
0.1002
3 3
0.1017
3 4
0.0475
4 1
0.0009
4 2
0.0248
4 3
0.0684
4 4
0.0867