I'm trying to run an SEM, but I keep getting the following error message.
In sem.default(ram = ram, S = S, N = N, param.names = pars, var.names = vars, :
Could not compute QR decomposition of Hessian.
Optimization probably did not converge.
I have 4 latent variables (plant, AMF, abiotic, and soilAgg) with 2 or 4
indicator variables for each latent variable.My model is specified as:
> mod.AL.5 <-
specify.model()
1: plant ->
tnCSG, plantCSG, .6
2: plant ->
tnDIV, plantDIV, NA
3: plant ->
tnFRL, plantFRL, NA
4: plant ->
tnCRL, plantCRL, NA
5: AMF -> tnHL,
AMFHL, .6
6: AMF -> tnHRL,
AMFHRL, NA
7: abiotic ->
tnHYDRO, abioticHYDRO, .6
8: abiotic ->
tnPS, abioticPS, NA
9: abiotic ->
tnNaPi, abioticNaPi, NA
10: abiotic ->
tnNaPo, abioticNaPo, NA
11: soilAgg ->
tnWSA2.4, soilAgg2.4, .6
12: soilAgg ->
tnWSA1.2, soilAgg1.2, NA
13: soilAgg ->
tnWSA0.5.1, soilAgg0.5.1, NA
14: soilAgg ->
tnWSA212.500, soilAgg212.500, NA
15: plant <->
AMF, plantAMF, NA
16: abiotic
<-> plant, abioticPlant, NA
17: AMF <->
abiotic, AMFabiotic, NA
18: plant ->
soilAgg, plantAgg, NA
19: AMF ->
soilAgg, AMFagg, NA
20: abiotic ->
soilAgg, abioticAgg, NA
21: tnCSG <->
tnCSG, CSGerror, NA
22: tnDIV <->
tnDIV, DIVerror, NA
23: tnFRL <->
tnFRL, FRLerror, NA
24: tnCRL <->
tnCRL, CRLerror, NA
25: tnHL <->
tnHL, HLerror, NA
26: tnHRL <->
tnHRL, HRLerror, NA
27: tnHYDRO
<-> tnHYDRO, hydroError, NA
28: tnPS <->
tnPS, PSerror, NA
29: tnNaPi <->
tnNaPi, NaPiError, NA
30: tnNaPo <->
tnNaPo, NaPoError, NA
31: tnWSA2.4
<-> tnWSA2.4, WSA2.4Error, NA
32: tnWSA1.2
<-> tnWSA1.2, WSA1.2Error, NA
33: tnWSA0.5.1
<-> tnWSA0.5.1, WSA0.5.1Error, NA
34: tnWSA212.500
<-> tnWSA212.500, WSA212.500Error, NA
35: plant <->
plant, NA, 1
36: AMF <->
AMF, NA, 1
37: abiotic
<-> abiotic, NA, 1
38: soilAgg <->
soilAgg, NA, 1
39:
and here's the sem function I used.> sem.AL.5 <- sem(mod.AL.5, ALcor, N=200)
I've been through the examples in the SEM documentation, and the Wheaton
model (which I can run) is similar to mine. I found similar error messages in
the help archives, so I tried setting the error of the latent variables to 1,
and suggesting start values for one of the indicator variables associated with
each latent variable, but I still get the same error message. Does anyone have
any idea how to take care of this? Just from looking at the correlation matrix
(see below), the correlations are generally low. Would that be enough to prevent
convergence?
If I use the covariance matrix instead of the correlation matrix and set
maxiter=2000 I get output, but I still get an error message saying the maximum
number of iterations was exceeded and that optimization may not have converged.
Thanks in advance for your help,
Kathryn
> ALcor
tnCSG tnDIV tnFRL tnCRL tnHL
tnCSG 1.00000000 0.598589240 -0.09766885 -0.078573691 0.47290972
tnDIV 0.59858924 1.000000000 0.04533969 -0.001910876 0.26677920
tnFRL -0.09766885 0.045339692 1.00000000 0.853052765 0.23436161
tnCRL -0.07857369 -0.001910876 0.85305276 1.000000000 0.33387717
tnHL 0.47290972 0.266779205 0.23436161 0.333877174 1.00000000
tnHRL 0.30636700 0.179672277 -0.81941675 -0.729226499 -0.13339807
tnHYDRO -0.11891511 -0.025275225 0.35393818 0.267298106 0.25214845
tnNaPi -0.41824708 -0.736959179 0.14473067 0.249238485 -0.02147190
tnNaPo 0.10707415 -0.360450067 0.03880794 0.163099408 0.33992466
tnPS 0.27726142 0.510986276 0.07216890 0.083311567 0.40554097
tnWSA2.4 0.16649209 -0.057519363 0.04794688 -0.072041701 0.05880872
tnWSA1.2 -0.10414538 -0.205197511 -0.06031893 0.159312072 0.18604414
tnWSA0.5.1 -0.17479942 0.106380104 0.02144471 0.054702063 -0.16966552
tnWSA212.500 -0.11611759 0.169063525 -0.09266492 -0.084036868 -0.17802964
tnHRL tnHYDRO tnNaPi tnNaPo tnPS
tnCSG 0.30636700 -0.11891511 -0.41824708 0.107074151 0.27726142
tnDIV 0.17967228 -0.02527523 -0.73695918 -0.360450067 0.51098628
tnFRL -0.81941675 0.35393818 0.14473067 0.038807940 0.07216890
tnCRL -0.72922650 0.26729811 0.24923848 0.163099408 0.08331157
tnHL -0.13339807 0.25214845 -0.02147190 0.339924658 0.40554097
tnHRL 1.00000000 -0.16550993 -0.16140371 -0.055624417 0.02845425
tnHYDRO -0.16550993 1.00000000 0.17276186 0.256247649 0.25005594
tnNaPi -0.16140371 0.17276186 1.00000000 0.533883552 -0.40849332
tnNaPo -0.05562442 0.25624765 0.53388355 1.000000000 0.13696445
tnPS 0.02845425 0.25005594 -0.40849332 0.136964451 1.00000000
tnWSA2.4 -0.12351999 -0.16219698 -0.16570385 -0.002183605 -0.27257633
tnWSA1.2 0.07826866 0.02910394 0.44637209 0.285468889 -0.05344273
tnWSA0.5.1 0.09160446 0.17053215 0.10932858 -0.107123596 0.21283469
tnWSA212.500 0.16099662 0.14943925 -0.04625665 -0.182579243 0.27829650
tnWSA2.4 tnWSA1.2 tnWSA0.5.1 tnWSA212.500
tnCSG 0.166492091 -0.10414538 -0.17479942 -0.11611759
tnDIV -0.057519363 -0.20519751 0.10638010 0.16906352
tnFRL 0.047946878 -0.06031893 0.02144471 -0.09266492
tnCRL -0.072041701 0.15931207 0.05470206 -0.08403687
tnHL 0.058808717 0.18604414 -0.16966552 -0.17802964
tnHRL -0.123519990 0.07826866 0.09160446 0.16099662
tnHYDRO -0.162196983 0.02910394 0.17053215 0.14943925
tnNaPi -0.165703851 0.44637209 0.10932858 -0.04625665
tnNaPo -0.002183605 0.28546889 -0.10712360 -0.18257924
tnPS -0.272576334 -0.05344273 0.21283469 0.27829650
tnWSA2.4 1.000000000 -0.51857209 -0.91353074 -0.83749935
tnWSA1.2 -0.518572095 1.00000000 0.24990500 0.03657620
tnWSA0.5.1 -0.913530744 0.24990500 1.00000000 0.87048428
tnWSA212.500 -0.837499354 0.03657620 0.87048428 1.00000000
[[alternative HTML version deleted]]