Running CVA/HB

Top  Previous  Next

Once you have collected conjoint data for a CVA project, you are ready to estimate part-worths with HB.  Click Analysis | Analysis Manager....  Click the Settings... icon gear_blue to review or change the default settings.


Estimation Settings:


Number of Iterations:  This describes how many "burn-in" iterations will be used, prior to assuming convergence.  10,000 iterations usually provides enough initial iterations so that the part-worth utilities "converge."  Convergence is assessed typically by viewing the visual display of the history of part-worth estimates.  If the lines in the display tend to be wobbling horizontally, with no noticeable trend, then this is evidence of convergence.  If the parameters do not seem to converge, then you can increase the number of initial iterations.


Number of Draws:  This describes how many "draws" or candidate betas will be used (averaged to form a point estimate) for each respondent.  10,000 draws tends to give quite good precision for each respondent.  However, you can obtain slightly better precision by increasing the number of used draws.


Prior Degrees of Freedom:  This value is the additional degrees of freedom for the prior covariance matrix (not including the # parameters to be estimated). The higher the value, the greater the influence of the prior variance and more data are needed to change that prior. The scaling for degrees of freedom is relative to the sample size. If you use 50 and you only have 100 subjects, then the prior will have a big impact on the results. If you have 1000 subjects, you will get about the same result if you use a prior of 5 or 50. As an example of an extreme case, with 100 respondents and a prior variance of 0.1 with prior degrees of freedom set to the number of parameters estimated plus 50, each respondent's resulting part-worths will vary relatively little from the population means. We urge users to be careful when setting the prior degrees of freedom, as large values (relative to sample size) can make the prior exert considerable influence on the results.  


Prior Variance:  The default is 2 for the prior variance for each parameter, but users can modify this value. Increasing the prior variance tends to place more weight on fitting each individual's data, and places less emphasis on "borrowing" information from the population parameters. The resulting posterior estimates are relatively insensitive to the prior variance, except 1) when there is very little information available within the unit of analysis relative to the number of estimated parameters, and 2) the prior degrees of freedom for the covariance matrix (described above) is relatively large.  


Random Starting Seed: HB requires a random number generator, and therefore a starting seed is required.  By default, the random seed is 1, but users can specify a specific seed to use (integers from 1 to 32000), so that results are repeatable. When using different random seeds, the posterior estimates will vary, but insignificantly, assuming convergence has been reached and many draws have been used.  If you use a seed of "0", a seed is drawn based on the computer's clock, so a different seed will be chosen if you repeat the analysis.




The Scale Settings (Minimum Scale, Scale Direction, and Recode Method) are similar to those described for OLS estimation





After the computation finishes, the new HB run is automatically saved for use in the choice simulator.



Page link: