Abstract
The evidence for a hypothesized link function in a generalized linear model may depend crucially on one or a few observations. Global and local influence diagnostics are proposed to assess influence on the goodness-of-link test. It is found that the local influence method can reveal influential observations on the link even when masking is present.
Original language | English |
---|---|
Pages (from-to) | 351-358 |
Number of pages | 8 |
Journal | Statistics and Probability Letters |
Volume | 31 |
Issue number | 4 |
DOIs |
|
Publication status | Published - 1997 |