Re-run simulations with various randomization strategies

marcdotson / conjoint-ensembles

Using clever randomization and ensembling strategies to accommodate multiple data pathologies in conjoint studies.

MIT License

0 stars 1 forks source link

Re-run simulations with various randomization strategies #43

Closed dgmiller closed 4 years ago

dgmiller commented 5 years ago

@marcdotson Randomize over

respondents
choice tasks
attributes and level
alternatives
joint randomization

marcdotson commented 5 years ago

Save the model output so we can test different measures of fit, including predictive fit for an actual hold-out sample of respondents and not just the hold-out task (though with simulation we might as well investigate both).

marcdotson commented 4 years ago

Combining common issues.