Latent GOLD Choice software description

LatentGOLD, even as SI-CHAID, and GOLDMineR, is a registered trademarks of Statistical Innovations Inc., Belmont MA. USA. Recently, this aggregate model has been improved to allow for the fact that different consumer segments utilize different preferences in making their choices. The result is a model that produces better share estimates by simultaneously identifying the important segments and the estimated share for each segment. Latent GOLD Choice® represents the GOLD standard for developing advanced choice models. Choice data is obtained from surveys or actual behavior where respondents rate/rank/choose products/services/alternatives/options. Choice models differ from traditional regression models in that choices are predicted as a function of characteristics of the choice alternatives. Each alternative/product/service/option has attributes. What is estimated is the importances/utilities of these attributes. Latent classes represent segments that give differential importance to the various attributes.

The two most popular ways to take into account differences in respondent preferences are Hierarchical Bayes (HB) models and Latent Class (LC) models, also know as finite mixture models. A recent extensive comparison of the two was made by Andrews, Ainslie and Currim, (2002), An empirical comparison of logit choice models with discrete vs. continuous representations of heterogeneity, Journal of Marketing Research, Vol. XXXIX (November), 479-487. In a followup publication by Andrews and Currim, (May 2003, JMR), the authors refer to their earlier work as ..showing that finite mixture [LC] models are at least as effective as more recent methods [HB] for recovering heterogeneity.. Added to the fact that the Latent GOLD Choice® program can estimate models in a fraction of the time that it takes to estimate HB models, plus provides many additional capabilities, we believe that Latent GOLD Choice® is the GOLD standard for advanced choice modeling

New Features in Latent GOLD Choice® 4.0

Known Class Indicator

This feature allows more control over the segment definitions by pre-assigning selected cases (not) to be in a particular class or classes.

For more information, see Tutorial #5: Using Latent GOLD® 4.0 with the Known Class Option. In this tutorial, we illustrate the use of the ‘known class’ feature in Latent GOLD® 4.0 to take into account additional information on a subset of cases which allows us to classify them into a particular class with probability one. In this case, the information comes from a physician’s diagnosis of the patient as `Depressed` or merely `Troubled`, corresponding to 2 of the 3 latent classes.

Conditional Bootstrap p-value

Model difference bootstrap can be used to formally assess the significance in improvement associated with adding additional classes, additional DFactors and/or an additional DFactor levels to the model, or to relax any other model restriction.

Overdispersed (Count and Binomial Count in Regression)

Overdispersion is a common phenomenon in count data. It means that, as a result of unobserved heterogeneity, the variance of the count variable is larger than estimated by the Poisson (binomial) model. The overdispersed option makes it possible to account for unobserved heterogeneity by assuming that the rates (success probabilities) follow a gamma (beta) distribution. This yields a negative-binomial model for overdispersed Poisson counts and a negative-binomial model for overdispersed binomial counts. Note that this option is conceptually similar to including a normally distributed random intercept in a regression model for a count variable.

The overdispersion option is useful if one wishes to analyze count data using mixture or zero-inflated variants of (truncated) negative-binomial or beta-binomial models (Agresti, 2000; Long, 1997; Simonoff, 2003). The negative-binomial model is a Poisson model with an extra error term coming from a gamma distribution. The beta-binomial model is a variant of the binomial count model that assumes that the success probabilities come from a beta distribution. These models are common in fields such as criminology, political sciences, medicine, biology, and marketing.

New Features in Latent GOLD Choice 4.0 Advanced Module

Continuous latent variables (CFactors)

An option for specifying models containing continuous latent variables, called CFactors, in a cluster, DFactor or regression model. CFactors can be used to specify continuous latent variable models, such as factor analysis and item response theory models, and regression models with continuous random effects. For more details, see:

- Popper, Richard, Kroll, Jeff and Magidson, Jay (2004).
Applications of latent class models to food product development: a case study
Sawtooth Software Proceedings, 2004.
- Tutorial #6: Estimating a Random Intercept Regression Model.

In this tutorial, we illustrate the use of continuous factors (CFactors) to control for the level effect in ratings data. A latent class regression model is estimated where the dependent variable is ratings of 15 crackers on taste, and 12 predictors correspond to different attributes of the crackers. Different classes are identified that show different taste preferences, controlling for their overall rating level. These data are based on a paper by Popper et. al. The use of CFactors requires the Advanced version of Latent GOLD® 4.0.

Multilevel Modeling

An option for defining two-level data variants of any model implemented in Latent GOLD. Group-level variation may be accounted for by specifying group-level latent classes (GClasses) and/or group-level CFactors (GCFactors). In addition, when 2 or more GClasses are specified, group-level covariates (GCovariates) can be included in the model to describe/predict them. The multilevel option can also be used for specifying three-level parametric or nonparametric random-effects regression models. Sumultaneously develop country-level and individual level segments. See:

- Bijmolt, T.H., Paas, L.J., Vermunt , J.K. (2004).
Country and Consumer Segmentation: Multi-level Latent Class Analysis of Financial Product Ownership, International Journal of Research in Marketing, 21, 323-340
- Vermunt, J.K, and Magidson, J. (2005).
Hierarchical mixture models for nested data structures, In C. Weihs und W. Gaul (eds), Classification: The Ubiquitous Challenge. Heidelberg: Springer.

Survey Options for complex sample data

Two important survey sampling designs are stratified sampling -- sampling cases within strata, and two-stage cluster sampling -- sampling within primary sampling units (PSUs) and subsequent sampling of cases within the selected PSUs. Moreover, sampling weights may exist. The Survey option takes the sampling design and the sampling weights into account when computing standard errors and related statistics associated with the parameter estimates, and estimates the design effect

SI-CHAID® 4.0 add-on to Latent GOLD Choice®

Whenever covariates are available to describe latent classes obtained from Latent GOLD® 4.0, the SI-CHAID® 4.0 add-on can provide an especially valuable alternative treatment to the use of active and/or inactive covariates in Latent GOLD® 4.0 under any of the following conditions:

- when many covariates are available and you wish to know which ones are most important
- when you do not wish to specify certain covariates as active because you do not wish them to affect the model parameters, but you still desire to assess their statistical significance with respect to the classes (or a specified subset of the classes)
- when you wish to develop a separate profile for each latent class
- when you wish to explore differences between 2 or more selected latent classes using a tree modeling structure
- when the relationship between the covariates and classes is nonlinear or includes interaction effects, or
- when you wish to profile order-restricted latent classes or discrete factors (Dfactors) - new in Latent GOLD® 4.0

This option is illustrated in the following tutorials:

- Latent GOLD® 4.0 Tutorial 4: Profiling LC Segments using the CHAID Option
- SI-CHAID® 4.0 Tutorial 4: Using CHAID with Multiple Correlated Dependent Variables
- Also, see this article: An Extension of the CHAID Tree-based Segmentation Algorithm to Multiple Dependent Variables, Magidson and Vermunt (2005)

Latent GOLD Syntax module

Discover the full power of the Advanced versions Latent GOLD® and Latent GOLD® Choice with the new LG-Syntax TM module. This new Syntax module will allow you to estimate many new models including hidden Markov models, score new data files, submit runs in batch-mode, impute missing data, include only selected cases and obtain new output. Do not miss it!

LG-SyntaxTM Module Features:

- Run in interactive or batch mode
- Use saved models to score new cases
- Simulate data from any model
- Perform power calculations
- Impute (fill in) missing values in input data file
- Obtain additional output without re-estimating model
- Powerful / more flexible control over parameter restrictions
- Seamless data fusion, regress with several dependent variables to develop
- hybrid LC Choice / rating models
- hybrid MaxDiff / rating models
- Scale-factor adjusted LC choice models
- Latent path models
- Hidden Markov models

Ordering

For price information and ordering, please visit the Prices and Ordering Page

Manufacturer page

Science Plus Group is a distributor for this product. You can also visit the Statistical Innovations website, the manufacturer of this product.