Mathematics
8 years ago

New testing methods improve accuracy of count regression models significantly.

Statistica Sinica
Testing for zero-modification in count regression models.
Aleksey Min, Claudia Czado

Paper Summary

Related papers

Theme background image
Mathematics
3 years ago

New method improves accuracy of statistical tests in data analysis.

Missing Data
Skew Distributions
Financial Volatility
1 view
Theme background image
Environmental Science
5 years ago

New test improves accuracy in studying species occurrence for ecology research.

Biodiversity Conservation
Pollinator Decline
Discrete Choice
1 view
Theme background image
Mathematics
5 years ago

New method improves accuracy of small-sample regression tests without extra computation!

Missing Data
Regularization Methods
Multicollinearity

Count data often have issues like overdispersion and zero outcomes. Researchers tested different methods for adjusting these issues in regression models. They found that using Wald and likelihood ratio tests is more powerful than score tests. In some cases, the score test resulted in a power loss of up to 87% compared to the Wald test. This means that using Wald and likelihood ratio tests is more effective in these situations. The researchers also showed that the maximum likelihood estimates in certain regression models are consistent and asymptotically normal, which is important for the Wald and likelihood ratio tests to work. Overall, these findings highlight the importance of using the most powerful testing methods for count regression models.