Accepted Solutions
Hi @Syoshinnsya2024,
Welcome in the Community !
Did you already check the JMP Help section related to Choice models ? : Overview of the Choice Platform
"The Choice platform uses a form of conditional logistic regression to estimate the probability that a configuration is preferred. Unlike simple logistic regression, choice modeling uses a linear model to model choices based on response attributes and not solely upon subject characteristics."
Some articles/litterature are also provided to dive deeper in the understanding of Choice models :
- Louviere, J. J., Flynn, T. N., and Marley, A. A. (2015). Best-Worst Scaling: Theory, Methods and Applications. Cambridge: Cambridge University Press.
- Train, K. E. (2009). Discrete Choice Methods and Simulation. 2nd ed. Cambridge University Press.
- Rossi, P. E., Allenby, G. M., and McCulloch, R. (2005). Bayesian Statistics and Marketing. Chichester, UK: John Wiley & Sons.
As it is based on a type of logistic regression, you won't have access to "traditional" R².
For logistic regression, McFadden’s pseudo R² can be used.
For Choice models, you can use the different metrics related to Likelihood, like logLikelihhod, AICc and BIC metrics. There are however most useful when comparing multiple models, to choose the most informative and less complex one.
If you want to compute McFadden’s pseudo R² in your example, you could technically do it by using this formula :
With Lm : the log-likelihood of full model
and L0 : the log-likelihood of model with only intercept (add a constant factor value and use it in model effects).
In the example "Choice Profiles", I run a model with effects having a logLikelihood equal to -6 (or -2logLikelihood = 12).
When running a model with only a constant effect (intercept), I have a logLikelihood equal to -9 (or -2logLikelihood = 18).
Using the formula above, I can calculate a McFadden’s pseudo R² equal to 0,333.
Attached is the datatable used for the models so that you can reproduce the steps above.
Hope this answer will help you,
Hi @Syoshinnsya2024,
Welcome in the Community !
Did you already check the JMP Help section related to Choice models ? : Overview of the Choice Platform
"The Choice platform uses a form of conditional logistic regression to estimate the probability that a configuration is preferred. Unlike simple logistic regression, choice modeling uses a linear model to model choices based on response attributes and not solely upon subject characteristics."
Some articles/litterature are also provided to dive deeper in the understanding of Choice models :
- Louviere, J. J., Flynn, T. N., and Marley, A. A. (2015). Best-Worst Scaling: Theory, Methods and Applications. Cambridge: Cambridge University Press.
- Train, K. E. (2009). Discrete Choice Methods and Simulation. 2nd ed. Cambridge University Press.
- Rossi, P. E., Allenby, G. M., and McCulloch, R. (2005). Bayesian Statistics and Marketing. Chichester, UK: John Wiley & Sons.
As it is based on a type of logistic regression, you won't have access to "traditional" R².
For logistic regression, McFadden’s pseudo R² can be used.
For Choice models, you can use the different metrics related to Likelihood, like logLikelihhod, AICc and BIC metrics. There are however most useful when comparing multiple models, to choose the most informative and less complex one.
If you want to compute McFadden’s pseudo R² in your example, you could technically do it by using this formula :
With Lm : the log-likelihood of full model
and L0 : the log-likelihood of model with only intercept (add a constant factor value and use it in model effects).
In the example "Choice Profiles", I run a model with effects having a logLikelihood equal to -6 (or -2logLikelihood = 12).
When running a model with only a constant effect (intercept), I have a logLikelihood equal to -9 (or -2logLikelihood = 18).
Using the formula above, I can calculate a McFadden’s pseudo R² equal to 0,333.
Attached is the datatable used for the models so that you can reproduce the steps above.
Hope this answer will help you,