Notice: On April 23, 2014, Statalist moved from an email list to a forum, based at statalist.org.
[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
Re: st: Odd SEM Results
From
Joseph Trubisz <[email protected]>
To
"[email protected]" <[email protected]>
Subject
Re: st: Odd SEM Results
Date
Mon, 05 Aug 2013 15:36:57 -0400
I sent this to Stata to see if they can find the issue. I showed it to a colleague and he is stumped as well. He's actually concerned because he's not sure if some of his prior results are correct. Will post the solution if I get one.
Joe Trubisz
(516) 729-7559 (cell)
On Aug 5, 2013, at 3:30 PM, Anders Alexandersson <[email protected]> wrote:
> Can you simplify the model (e.g., drop all the covariates), put
> everything in a reproducible do-file, and still get different results?
> That would greatly simplify the problem. If you provide a reproducible
> do-file, I and maybe others would try it and report back.
> Alternatively, you could contact StataCorp since -sem- is an official
> command and therefore supported.
>
> Anders Alexandersson
> [email protected]
>
>
> On Mon, Aug 5, 2013 at 1:46 PM, Joseph Trubisz <[email protected]> wrote:
>> Tried that as well.
>> Exact same error.
>>
>> Joe
>>
>>
>> On Aug 5, 2013, at 11:13 AM, Anders Alexandersson <[email protected]> wrote:
>>
>>> Hi Joe, only the first model from the SEM builder specifies the
>>> options "latent(Intercept Slope ) nocapslatent" which, I guess, might
>>> be why this model has chi2(45) instead of chi2(43) and no output for
>>> _cons in the structural model. What happens if you instead use the
>>> default?
>>>
>>> Anders Alexandersson
>>> [email protected]
>>>
>>> On Sat, Aug 3, 2013 at 12:38 PM, Joseph Trubisz <[email protected]> wrote:
>>>> Greetings...
>>>>
>>>> I probably am just missing something, but I don't know what.
>>>> I'm attempting to use sembuilder to create the diagram from Acock's SEM book, specifically the example
>>>> as shown on p.188.
>>>>
>>>> If I use sembuilder, it generates the following output:
>>>>
>>>> . sem (Intercept@1 -> bmi01) (Intercept@1 -> bmi02) (Intercept@1 -> bmi03) (Intercept@1 -> bmi05) (Interc
>>>>> ept@1 -> bmi06) (Intercept@1 -> bmi07) (Intercept@1 -> bmi08) (Intercept@1 -> bmi09) (Slope@0 -> bmi01)
>>>>> (Slope@1 -> bmi02) (Slope@2 -> bmi03) (Slope@4 -> bmi05) (Slope@5 -> bmi06) (Slope@6 -> bmi07) (Slope@
>>>>> 7 -> bmi08) (Slope@8 -> bmi09) (_cons -> Intercept) (_cons -> Slope) (male -> Intercept) (male -> Slope
>>>>> ) (wgtc -> Intercept) (wgtc -> Slope) if bmi01!=.|bmi02!=.|bmi03!=.|bmi05!=.|bmi06!=.|bmi07!=.|bmi08!=.
>>>>> |bmi09!=., method(mlmv) latent(Intercept Slope ) var( e.Intercept*e.Slope) nocapslatent noconstant
>>>> note: Missing values found in observed exogenous variables. Using the noxconditional behavior. Specify
>>>> the forcexconditional option to override this behavior.
>>>> Endogenous variables
>>>>
>>>> Measurement: bmi01 bmi02 bmi03 bmi05 bmi06 bmi07 bmi08 bmi09
>>>> Latent: Intercept Slope
>>>>
>>>> Exogenous variables
>>>>
>>>> Observed: male wgtc
>>>>
>>>> Fitting saturated model:
>>>>
>>>> Iteration 0: log likelihood = -30162.223
>>>> Iteration 1: log likelihood = -29297.714
>>>> Iteration 2: log likelihood = -28707.587
>>>> Iteration 3: log likelihood = -28564.929
>>>> Iteration 4: log likelihood = -28557.353
>>>> Iteration 5: log likelihood = -28557.204
>>>> Iteration 6: log likelihood = -28557.204
>>>>
>>>> Fitting baseline model:
>>>>
>>>> Iteration 0: log likelihood = -36523.419
>>>> Iteration 1: log likelihood = -36520.845
>>>> Iteration 2: log likelihood = -36520.836
>>>> Iteration 3: log likelihood = -36520.836
>>>>
>>>> Fitting target model:
>>>>
>>>> Iteration 0: log likelihood = -52919.48 (not concave)
>>>> Iteration 1: log likelihood = -52675.161 (not concave)
>>>> Iteration 2: log likelihood = -52171.219 (not concave)
>>>> Iteration 3: log likelihood = -49397.835 (not concave)
>>>> Iteration 4: log likelihood = -42220.623 (not concave)
>>>> Iteration 5: log likelihood = -39274.796
>>>> Iteration 6: log likelihood = -38652.54
>>>> Iteration 7: log likelihood = -34772.666
>>>> Iteration 8: log likelihood = -32169.128
>>>> Iteration 9: log likelihood = -31367.639
>>>> Iteration 10: log likelihood = -30934.922
>>>> Iteration 11: log likelihood = -30910.018
>>>> Iteration 12: log likelihood = -30909.236
>>>> Iteration 13: log likelihood = -30909.234
>>>>
>>>> Structural equation model Number of obs = 1581
>>>> Estimation method = mlmv
>>>> Log likelihood = -30909.234
>>>>
>>>> ( 1) [bmi01]Intercept = 1
>>>> ( 2) [bmi02]Intercept = 1
>>>> ( 3) [bmi02]Slope = 1
>>>> ( 4) [bmi03]Intercept = 1
>>>> ( 5) [bmi03]Slope = 2
>>>> ( 6) [bmi05]Intercept = 1
>>>> ( 7) [bmi05]Slope = 4
>>>> ( 8) [bmi06]Intercept = 1
>>>> ( 9) [bmi06]Slope = 5
>>>> (10) [bmi07]Intercept = 1
>>>> (11) [bmi07]Slope = 6
>>>> (12) [bmi08]Intercept = 1
>>>> (13) [bmi08]Slope = 7
>>>> (14) [bmi09]Intercept = 1
>>>> (15) [bmi09]Slope = 8
>>>> (16) [bmi01]_cons = 0
>>>> (17) [bmi02]_cons = 0
>>>> (18) [bmi03]_cons = 0
>>>> (19) [bmi05]_cons = 0
>>>> (20) [bmi06]_cons = 0
>>>> (21) [bmi07]_cons = 0
>>>> (22) [bmi08]_cons = 0
>>>> (23) [bmi09]_cons = 0
>>>> --------------------------------------------------------------------------------
>>>> | OIM
>>>> | Coef. Std. Err. z P>|z| [95% Conf. Interval]
>>>> ---------------+----------------------------------------------------------------
>>>> Structural |
>>>> Intercept <- |
>>>> male | 26.90551 .6496371 41.42 0.000 25.63225 28.17878
>>>> wgtc | 6.996785 .6003745 11.65 0.000 5.820072 8.173497
>>>> -------------+----------------------------------------------------------------
>>>> Slope <- |
>>>> male | .3657208 .02104 17.38 0.000 .3244831 .4069584
>>>> wgtc | .0889063 .0194718 4.57 0.000 .0507423 .1270703
>>>> ---------------+----------------------------------------------------------------
>>>> Measurement |
>>>> bmi01 <- |
>>>> Intercept | 1 (constrained)
>>>> _cons | 0 (constrained)
>>>> -------------+----------------------------------------------------------------
>>>> bmi02 <- |
>>>> Intercept | 1 (constrained)
>>>> Slope | 1 (constrained)
>>>> _cons | 0 (constrained)
>>>> -------------+----------------------------------------------------------------
>>>> bmi03 <- |
>>>> Intercept | 1 (constrained)
>>>> Slope | 2 (constrained)
>>>> _cons | 0 (constrained)
>>>> -------------+----------------------------------------------------------------
>>>> bmi05 <- |
>>>> Intercept | 1 (constrained)
>>>> Slope | 4 (constrained)
>>>> _cons | 0 (constrained)
>>>> -------------+----------------------------------------------------------------
>>>> bmi06 <- |
>>>> Intercept | 1 (constrained)
>>>> Slope | 5 (constrained)
>>>> _cons | 0 (constrained)
>>>> -------------+----------------------------------------------------------------
>>>> bmi07 <- |
>>>> Intercept | 1 (constrained)
>>>> Slope | 6 (constrained)
>>>> _cons | 0 (constrained)
>>>> -------------+----------------------------------------------------------------
>>>> bmi08 <- |
>>>> Intercept | 1 (constrained)
>>>> Slope | 7 (constrained)
>>>> _cons | 0 (constrained)
>>>> -------------+----------------------------------------------------------------
>>>> bmi09 <- |
>>>> Intercept | 1 (constrained)
>>>> Slope | 8 (constrained)
>>>> _cons | 0 (constrained)
>>>> ---------------+----------------------------------------------------------------
>>>> Mean |
>>>> male | .4990512 .0125749 39.69 0.000 .474405 .5236975
>>>> wgtc | -.0001655 .0192488 -0.01 0.993 -.0378925 .0375614
>>>> ---------------+----------------------------------------------------------------
>>>> Variance |
>>>> e.bmi01 | 2.618815 .2046736 2.246876 3.052323
>>>> e.bmi02 | 4.086149 .2175673 3.681222 4.535619
>>>> e.bmi03 | 4.674361 .2320231 4.241024 5.151974
>>>> e.bmi05 | 5.778033 .2604252 5.289505 6.311681
>>>> e.bmi06 | 8.181968 .3511234 7.521926 8.899928
>>>> e.bmi07 | 3.794672 .1906747 3.438769 4.187409
>>>> e.bmi08 | 2.92201 .1689406 2.608965 3.272618
>>>> e.bmi09 | 3.308288 .2088301 2.923295 3.743984
>>>> e.Intercept | 324.2532 11.61219 302.2741 347.8304
>>>> e.Slope | .2448578 .0117764 .2228311 .269062
>>>> male | .2499991 .0088918 .2331651 .2680484
>>>> wgtc | .5851282 .0208269 .5456996 .6274057
>>>> ---------------+----------------------------------------------------------------
>>>> Covariance |
>>>> e.Intercept |
>>>> e.Slope | 4.109434 .280162 14.67 0.000 3.560327 4.658542
>>>> -------------+----------------------------------------------------------------
>>>> male |
>>>> wgtc | -.0751429 .0098082 -7.66 0.000 -.0943666 -.0559193
>>>> --------------------------------------------------------------------------------
>>>> LR test of model vs. saturated: chi2(45) = 4704.06, Prob > chi2 = 0.0000
>>>>
>>>>
>>>> However, the output is nothing like what's in the book. However, if I type in
>>>> exactly what's in the book (p.188), I get the correct results as shown below:
>>>>
>>>> . sem (Intercept@1 Slope@0->bmi01) (Intercept@1 Slope@1->bmi02) (Intercept@1 Slope@2->bmi03) (Intercept@1
>>>>> Slope@4->bmi05)(Intercept@1 Slope@5->bmi06)(Intercept@1 Slope@6->bmi07)(Intercept@1 Slope@7->bmi08)(In
>>>>> tercept@1 Slope@8->bmi09) (Intercept Slope<-male wgtc _cons) if bmi01!=.|bmi02!=.|bmi03!=.|bmi05!=.|bmi
>>>>> 06!=.|bmi07!=.|bmi08!=.|bmi09!=.,var(e.Intercept*e.Slope) method(mlmv) noconstant
>>>> note: Missing values found in observed exogenous variables. Using the noxconditional behavior. Specify
>>>> the forcexconditional option to override this behavior.
>>>> Endogenous variables
>>>>
>>>> Measurement: bmi01 bmi02 bmi03 bmi05 bmi06 bmi07 bmi08 bmi09
>>>> Latent: Intercept Slope
>>>>
>>>> Exogenous variables
>>>>
>>>> Observed: male wgtc
>>>>
>>>> Fitting saturated model:
>>>>
>>>> Iteration 0: log likelihood = -30162.223
>>>> Iteration 1: log likelihood = -29297.714
>>>> Iteration 2: log likelihood = -28707.587
>>>> Iteration 3: log likelihood = -28564.929
>>>> Iteration 4: log likelihood = -28557.353
>>>> Iteration 5: log likelihood = -28557.204
>>>> Iteration 6: log likelihood = -28557.204
>>>>
>>>> Fitting baseline model:
>>>>
>>>> Iteration 0: log likelihood = -36523.419
>>>> Iteration 1: log likelihood = -36520.845
>>>> Iteration 2: log likelihood = -36520.836
>>>> Iteration 3: log likelihood = -36520.836
>>>>
>>>> Fitting target model:
>>>>
>>>> Iteration 0: log likelihood = -52919.48 (not concave)
>>>> Iteration 1: log likelihood = -52663.873 (not concave)
>>>> Iteration 2: log likelihood = -52499.164 (not concave)
>>>> Iteration 3: log likelihood = -52371.927 (not concave)
>>>> Iteration 4: log likelihood = -46362.021 (not concave)
>>>> Iteration 5: log likelihood = -34630.285 (not concave)
>>>> Iteration 6: log likelihood = -34303.836 (not concave)
>>>> Iteration 7: log likelihood = -29724.362
>>>> Iteration 8: log likelihood = -29095.8
>>>> Iteration 9: log likelihood = -28787.969
>>>> Iteration 10: log likelihood = -28750.647
>>>> Iteration 11: log likelihood = -28750.02
>>>> Iteration 12: log likelihood = -28750.019
>>>>
>>>> Structural equation model Number of obs = 1581
>>>> Estimation method = mlmv
>>>> Log likelihood = -28750.019
>>>>
>>>> ( 1) [bmi01]Intercept = 1
>>>> ( 2) [bmi02]Intercept = 1
>>>> ( 3) [bmi02]Slope = 1
>>>> ( 4) [bmi03]Intercept = 1
>>>> ( 5) [bmi03]Slope = 2
>>>> ( 6) [bmi05]Intercept = 1
>>>> ( 7) [bmi05]Slope = 4
>>>> ( 8) [bmi06]Intercept = 1
>>>> ( 9) [bmi06]Slope = 5
>>>> (10) [bmi07]Intercept = 1
>>>> (11) [bmi07]Slope = 6
>>>> (12) [bmi08]Intercept = 1
>>>> (13) [bmi08]Slope = 7
>>>> (14) [bmi09]Intercept = 1
>>>> (15) [bmi09]Slope = 8
>>>> (16) [bmi01]_cons = 0
>>>> (17) [bmi02]_cons = 0
>>>> (18) [bmi03]_cons = 0
>>>> (19) [bmi05]_cons = 0
>>>> (20) [bmi06]_cons = 0
>>>> (21) [bmi07]_cons = 0
>>>> (22) [bmi08]_cons = 0
>>>> (23) [bmi09]_cons = 0
>>>> --------------------------------------------------------------------------------
>>>> | OIM
>>>> | Coef. Std. Err. z P>|z| [95% Conf. Interval]
>>>> ---------------+----------------------------------------------------------------
>>>> Structural |
>>>> Intercept <- |
>>>> male | 1.555545 .2436739 6.38 0.000 1.077953 2.033137
>>>> wgtc | 3.759441 .160021 23.49 0.000 3.445806 4.073077
>>>> _cons | 24.85781 .170538 145.76 0.000 24.52356 25.19206
>>>> -------------+----------------------------------------------------------------
>>>> Slope <- |
>>>> male | .0173321 .0271012 0.64 0.522 -.0357853 .0704495
>>>> wgtc | .0459197 .0178505 2.57 0.010 .0109333 .0809062
>>>> _cons | .3430045 .0189564 18.09 0.000 .3058506 .3801584
>>>> ---------------+----------------------------------------------------------------
>>>> Measurement |
>>>> bmi01 <- |
>>>> Intercept | 1 (constrained)
>>>> _cons | 0 (constrained)
>>>> -------------+----------------------------------------------------------------
>>>> bmi02 <- |
>>>> Intercept | 1 (constrained)
>>>> Slope | 1 (constrained)
>>>> _cons | 0 (constrained)
>>>> -------------+----------------------------------------------------------------
>>>> bmi03 <- |
>>>> Intercept | 1 (constrained)
>>>> Slope | 2 (constrained)
>>>> _cons | 0 (constrained)
>>>> -------------+----------------------------------------------------------------
>>>> bmi05 <- |
>>>> Intercept | 1 (constrained)
>>>> Slope | 4 (constrained)
>>>> _cons | 0 (constrained)
>>>> -------------+----------------------------------------------------------------
>>>> bmi06 <- |
>>>> Intercept | 1 (constrained)
>>>> Slope | 5 (constrained)
>>>> _cons | 0 (constrained)
>>>> -------------+----------------------------------------------------------------
>>>> bmi07 <- |
>>>> Intercept | 1 (constrained)
>>>> Slope | 6 (constrained)
>>>> _cons | 0 (constrained)
>>>> -------------+----------------------------------------------------------------
>>>> bmi08 <- |
>>>> Intercept | 1 (constrained)
>>>> Slope | 7 (constrained)
>>>> _cons | 0 (constrained)
>>>> -------------+----------------------------------------------------------------
>>>> bmi09 <- |
>>>> Intercept | 1 (constrained)
>>>> Slope | 8 (constrained)
>>>> _cons | 0 (constrained)
>>>> ---------------+----------------------------------------------------------------
>>>> Mean |
>>>> male | .4990512 .0125749 39.69 0.000 .474405 .5236975
>>>> wgtc | -.0009276 .0192434 -0.05 0.962 -.0386439 .0367888
>>>> ---------------+----------------------------------------------------------------
>>>> Variance |
>>>> e.bmi01 | 2.443884 .191981 2.095144 2.850673
>>>> e.bmi02 | 4.1472 .2157233 3.745229 4.592315
>>>> e.bmi03 | 4.772852 .2325326 4.33818 5.251077
>>>> e.bmi05 | 5.789807 .2596509 5.302625 6.321749
>>>> e.bmi06 | 8.228898 .3520222 7.56708 8.948599
>>>> e.bmi07 | 3.810727 .1909167 3.454322 4.203904
>>>> e.bmi08 | 2.922193 .1687984 2.609396 3.272487
>>>> e.bmi09 | 3.298484 .2089123 2.913418 3.734444
>>>> e.Intercept | 20.5145 .8043478 18.99706 22.15315
>>>> e.Slope | .189059 .0097265 .170925 .2091168
>>>> male | .2499991 .0088918 .2331651 .2680484
>>>> wgtc | .5849257 .0208131 .5455228 .6271748
>>>> ---------------+----------------------------------------------------------------
>>>> Covariance |
>>>> e.Intercept |
>>>> e.Slope | -.0272943 .0635517 -0.43 0.668 -.1518533 .0972647
>>>> -------------+----------------------------------------------------------------
>>>> male |
>>>> wgtc | -.0747626 .0098037 -7.63 0.000 -.0939775 -.0555478
>>>> --------------------------------------------------------------------------------
>>>> LR test of model vs. saturated: chi2(43) = 385.63, Prob > chi2 = 0.0000
>>>>
>>>> Problem: I look at the command not working and comparing it to the command that does
>>>> work and I don't see the difference.
>>>>
>>>> Can anyone point out to me where I might be going wrong?
>>> *
>>> * For searches and help try:
>>> * http://www.stata.com/help.cgi?search
>>> * http://www.stata.com/support/faqs/resources/statalist-faq/
>>> * http://www.ats.ucla.edu/stat/stata/
>>
>>
>> *
>> * For searches and help try:
>> * http://www.stata.com/help.cgi?search
>> * http://www.stata.com/support/faqs/resources/statalist-faq/
>> * http://www.ats.ucla.edu/stat/stata/
> *
> * For searches and help try:
> * http://www.stata.com/help.cgi?search
> * http://www.stata.com/support/faqs/resources/statalist-faq/
> * http://www.ats.ucla.edu/stat/stata/
*
* For searches and help try:
* http://www.stata.com/help.cgi?search
* http://www.stata.com/support/faqs/resources/statalist-faq/
* http://www.ats.ucla.edu/stat/stata/