Why does gamlss give incorrect estimates of ex-gaussian distribution parameters?

deedis 注册会员
2023-01-26 03:45

The estimates for sigma and nu do appear to be way off in the model output. This is because exGaus() uses a log link for both those parameters by default. From the documentation, showing the defaults:

The output shows results on the model scale so the estimates for those two parameters re given on the log scale.

If we exponentiate the results for sigma and nu then we get estimates that look more reasonable and much closer to what we'd expect.

# estimated sigma should be close to 35
#> [1] 32.98325

# estimated nu shoud be close to 100
#> [1] 102.5141

Created on 2021-10-11 by the reprex package (v2.0.0)