Speaker
Prof.
André Luis Santos de pinho
(Universidade Federal do Rio Grande do Norte)
Description
It is common to use model performance measures, such as AIC and BIC, to evaluate how well the model fits the data. This work illustrates that we need to go beyond these measures to assess a model's capability to represent the data. There are several ways to achieve that. Here we focus on a graphical approach using the probability integral transform (PIT) histogram. We present a situation in time series with non-negative integer values with more ones than the probabilistic model was able to explain.
| Keywords | integer valued time series, probability distribution, Engineering Statistics |
|---|
Primary authors
Prof.
André Luis Santos de pinho
(Universidade Federal do Rio Grande do Norte)
Prof.
LUZ MILENA ZEA FERNANDEZ
(Universidade Federal do Rio Grande do Norte)
Ms
BEATRIZ ARIADNA DA SILVA CIRIACO
(Universidade Federal do Rio Grande do Norte)