Message/Author 


Dear Linda, Owing to the computational demanding on the analysis I have to reduce the number of integration to 7, and it works. But, I would like to know if this has compromised the obtained results, i.e. if the results are well valid. Will there be any particular areas that I should pay attention to when I read the results? Thanks. Pat 


The only way to know is to use more integration points and see if you obtain the same results. How many dimensions of integration do you have? 


Dear Linda, Thanks for the advise. The coefficients obtained for the model (3 dimensions of integration) with default integration and with 7 integration are generally the same, although not identical. I will need 57 dimensions of integration in the full model. Pat 


You can also try MONTECARLO = 5000; 


Thanks, Linda. I have tried Integration = Montecarlo (5000) in the analysis with 5 Xvariables or dimensions of integration. I suspect that this is the way to control the maximum number of total integration points in the analysis regardless the number of Xvariables I put into the model. However, it turned out that there is a great significant difference in results obtained from the analysis with Integration = Montecarlo (5000); and those with Integration = 7. For the latter, the total number of integration points is 16807. What would you suggest? Thanks. Pat 


If you must bring the covariates into the model because of missing data, I would suggest multiple imputation. 

Katerina Gk posted on Thursday, September 26, 2013  8:13 am



Dear Linda The next algorithm represent Y=i+aX+bM+cXM+E Transformational leadership (X)(par), job satisfaction (Y)(er), selfefficacy(M)as moderator(a). TITLE:project DATA:FILE="....dat"; NOBSERVATIONS=640; ANALYSIS: TYPE= RANDOM; ALGORITHM=INTEGRATION; integration = montecarlo; VARIABLE: NAMES= l1 l2 l3 ....; USEVARIABLES = l1 l2 l3... ; Missing are all (999); MODEL: er1 by ...; er2 by ...; er3 by ..; er4 by ..; er5 by ..; a1 by ...; a2 by ...; a3 by ...; par1 by... ; par2 by... ; par3 by...; par4 by...; par5 by...; par6 by ...; par by par1 par2 par3 par4 par5 par6; er1 er2 er3 er4 er5 ON par ; a1 a2 a3 ON par; indpara1  par XWITH a1 ; indpara2  par XWITH a2 ; indpara3  par XWITH a3 ; er1 ON indpara1; er1 ON indpara2 ; er1 ON indpara3 ; er2 ON indpara1; ... er3 ON indpara1; ... er4 ON indpara1 ; ... er5 ON indpara1 ; ... OUTPUT: stdyx ; I would like to know if it is correct the code looking the equation, and also is it normal the long time that needs to complete? Thanks in advance for your help! Katerina University of Thessaly(Greece) 


I don't see any obvious problems. Your model has three dimensions of integration which can be computationally demanding. I would run one latent variable interaction at a time. It is likely that they are not all signficant. 

Katerina Gk posted on Thursday, September 26, 2013  2:20 pm



Dear Linda Thank you for your message, I run only one interaction indpara1  par_b XWITH a1_w ; er1_w ON indpara1; er2_w ON indpara1; er3_w ON indpara1; er4_w ON indpara1 ; er5_w ON indpara1 ; but I get this error: THE ESTIMATED COVARIANCE MATRIX COULD NOT BE INVERTED. COMPUTATION COULD NOT BE COMPLETED IN ITERATION 463.CHANGE YOUR MODEL AND/OR STARTING VALUES. THE MODEL ESTIMATION DID NOT TERMINATE NORMALLY DUE TO AN ERROR IN THE COMPUTATION. CHANGE YOUR MODEL AND/OR STARTING VALUES. What I should check or change? Thank you for your help Katerina 


Please send data, input, output and license number to Support. 

Katerina Gk posted on Friday, September 27, 2013  7:47 am



I ve already sent it!!! Thank you very much!!! Katerina 

AT Jothees posted on Saturday, May 20, 2017  7:24 am



Hi, i am trying to run second order growth mixed model but my computer takes more than 10 hours to perform the analysis. At times it doesnt produce any output. Is there a way to speed the analysis ? Number of dimensions: 3 ; total number of integration points : 3375. Regards, Jo 


Please send the input, data set, and your license number to support@statmodel.com. 

Back to top 