Parameter
estimation
Minimizer, stepping,
local covariance
MCMC or nested sampling
–> Buchner+14
Stuck in local optima, non-gaussian optima, many parameters
Combining
multiple observations
Stacking the data
Hierarchical model –> Buchner+17a
Average may not be meaningful
Redshift blur
Goodness of fit
chi² test
Visualize!
Monte Carlo simulations
Model comparison
Invalid if at parameter border, model non-linear or not in the high-data regime (i.e., never valid)
Model comparison
Likelihood ratio, F-test
Bayesian evidence or likelihood ratio,
calibrate with Monte Carlo simulations
–> Buchner+14
Invalid if at parameter border, model non-linear or not in the high-data regime (i.e., never valid)