Convergence issues using mixlogitwtp

[Amanda Tomasdotter]

Dear Statalisters,

I’m conducting a choice experiment with five attributes, were two attributes have four continuous levels, and two attributes have two levels included as a dummy, and one attribute had three levels where I recoded this into three dummies and using one level in my model. I also added an attribute for price ranging in four levels. I run the mixlogitwtp command by A.R. Hole to estimate my model in WTP space. I received a error message: Warning: Convergence not achieved.

I started by declaring my data to be panel data, since each individual (id) are given seven choice questions with two alternatives:

CODE:

cmset id question alternatives

OUTPUT:

Panel data: Panels id and time question
Case ID variable: _caseid
Alternatives variable: alternatives
Panel by alternatives variable: _panelaltid (strongly balanced)
Time variable: fråga, 1 to 7
Delta: 1 unit

Note: Data have been xtset.


Then generated negprice= (-1)*price, to be able to corporate it into the log normal distribution.

CODE:

list id _caseid choice x1 x2 x3 x4 x5 negprice if id==9

OUTPUT

id	_caseid	choice	x1	x2	x3	x4	x5	negprice
9	57	0	260	0	1	0	12	-90
9	57	1	330	1	0	0	28	-25
9	58	0	130	0	0	0	12	-130
9	58	1	130	1	1	0	7	-25
9	59	0	260	1	1	1	18	-65
9	59	1	330	0	0	0	12	-65
9	60	0	260	1	0	1	7	-25
9	60	1	260	0	0	0	28	-90
9	61	0	560	0	1	0	28	-25
9	61	1	560	0	0	1	28	-90
9	62	0	130	1	0	0	7	-90
9	62	1	130	0	1	1	12	-65
9	63	0	330	0	0	0	28	-65
9	63	1	330	1	1	0	18	-130

CODE:

mixlogitwtp choice, group(_caseid) price(negprice) id(id) rand( x1 x2 x3 x4 x5) nrep(1000) burn(100) technique(bhhh) iterate(1000)

Iteration 0: log likelihood = -669.98139
Iteration 1: log likelihood = -666.86419 (backed up)
Iteration 2: log likelihood = -662.24822
Iteration 3: log likelihood = -657.09689
Iteration 4: log likelihood = -656.78398
Iteration 5: log likelihood = -655.64698
Iteration 6: log likelihood = -654.85472
Iteration 7: log likelihood = -653.50478
Iteration 8: log likelihood = -653.3682
Iteration 9: log likelihood = -653.30287
Iteration 10: log likelihood = -653.28295
Iteration 11: log likelihood = -653.24267
Iteration 12: log likelihood = -653.22792
Iteration 13: log likelihood = -653.21128
Iteration 14: log likelihood = -653.19447
Iteration 15: log likelihood = -653.18469
Iteration 16: log likelihood = -653.17412
Iteration 17: log likelihood = -653.17223
Iteration 18: log likelihood = -653.15394
Iteration 19: log likelihood = -653.15157
Iteration 20: log likelihood = -653.11819
Iteration 21: log likelihood = -653.11664
Iteration 22: log likelihood = -653.09851
Iteration 23: log likelihood = -653.0953
Iteration 24: log likelihood = -653.05923
Iteration 25: log likelihood = -653.04855
Iteration 26: log likelihood = -653.04038
Iteration 27: log likelihood = -653.03079
Iteration 28: log likelihood = -653.02477
Iteration 29: log likelihood = -653.01427
Iteration 30: log likelihood = -653.00891
Iteration 31: log likelihood = -652.99678
Iteration 32: log likelihood = -652.99036
Iteration 33: log likelihood = -652.97525
Iteration 34: log likelihood = -652.96686
Iteration 35: log likelihood = -652.96144
Iteration 36: log likelihood = -652.96081
Iteration 37: log likelihood = -652.64558
Iteration 38: log likelihood = -652.59462
Iteration 39: log likelihood = -652.55809
Iteration 40: log likelihood = -652.54935
Iteration 41: log likelihood = -652.53877
Iteration 42: log likelihood = -652.53268
Iteration 43: log likelihood = -652.52968
Iteration 44: log likelihood = -652.52941
Iteration 45: log likelihood = -652.52768 (backed up)
Iteration 46: log likelihood = -652.48455
Iteration 47: log likelihood = -652.47941
Iteration 48: log likelihood = -652.47597
Iteration 49: log likelihood = -652.4758 (backed up)
Iteration 50: log likelihood = -652.47578 (backed up)
Iteration 51: log likelihood = -652.47576 (backed up)
Iteration 52: log likelihood = -652.47576 (backed up)
Iteration 53: log likelihood = -652.47532
Iteration 54: log likelihood = -652.47507 (backed up)
Iteration 55: log likelihood = -652.47437 (backed up)
Iteration 56: log likelihood = -652.47418 (backed up)
Iteration 57: log likelihood = -652.47411 (backed up)
Iteration 58: log likelihood = -652.47409 (backed up)
Iteration 59: log likelihood = -652.47408 (backed up)
Iteration 60: log likelihood = -652.47408 (backed up)
Iteration 61: log likelihood = -652.47408 (backed up)
,.....,
Iteration 1000: log likelihood = -652.47408 (backed up)
convergence not achieved
Mixed logit model in WTP space				Number of obs = 2,122
				Wald chi2(6) = 2140.99
Log likelihood = -652.47408				Prob > chi2 = 0.0000
		OPG
choice	Coefficient	std. err.	z	P>z	[95% conf.	interval]
Mean
x1	2.020137	.5630591	3.59	0.000	.9165613	3.123712
x2	-112.2809	26.868	-4.18	0.000	-164.9413	-59.62062
x3	193.0917	45.25788	4.27	0.000	104.3879	281.7955
x4	229.0887	42.99712	5.33	0.000	144.8159	313.3615
x5	9.790245	3.151583	3.11	0.002	3.613255	15.96723
negprice	-5.43188	.3157098	-17.21	0.000	-6.05066	-4.8131
SD
x1	.2036296	4.910859	0.04	0.967	-9.421478	9.828737
x2	-68.31679	40.64345	-1.68	0.093	-147.9765	11.34291
x3	55.66972	48.15007	1.16	0.248	-38.70268	150.0421
x4	175.8745	56.2576	3.13	0.002	65.61161	286.1374
x5	15.9116	4.383527	3.63	0.000	7.320043	24.50315
negprice	1.322363	.4069266	3.25	0.001	.5248016	2.119925
Warning: Convergence not achieved.
The sign of the estimated standard deviations is irrelevant: interpret them as being positive

I was wondering if my results could still be used, even if it did not converge, but I wanted to see if anybody had a solution to this problem?
Thank you,
Amanda