The concept of GLMs
, as a unified class of regression models for discrete and continuous variables, was first introduced by Nelder and Wedderburn in their seminal 1972 JRSS/A paper.
The analysis of data through GLMs
is quite flexible, since for one linear structure several models can be obtained depending on the proposed distribution for the error and the chosen function of connection.
Second, the structure of our GLMs
may not have decomposed temporal patterns in small mammal density or plant cover as effectively as a true repeat-measure model.
This occurs because the logarithmic link function is canonical for the Poisson error distribution, and it is a general property of GLMs
fitted with canonical link and including an intercept or factor (Nelder and Wedderburn 1972, ter Braak et al.
Generalized additive models (Hastie and Tibshirani, 1986, 1987) extend GLMs
by relaxing the linear-in-parameters assumption of the predictor index [Eta] with a sum of one-dimensional nonparametric functions of each explanatory variable so that [Eta](x) = [[Sigma].sub.p][f.sub.p]([x.sub.p]).
analysis with ROI as the dependent variable, and with order of market entry (OME), product dimension of quality (ProdQ), and the interaction of OME and ProdQ as the independent factors, yielded a significant overall model (F = 6.75; p [is less than] .001).
performed on the five separate performance measures indicated that both task and outcome interdependence significantly influence some measures of [TABULAR DATA FOR TABLE 4 OMITTED] performance, and a Kruskal-Wallis nonparametric analysis of variance on the mean ranks confirms this finding for the measure of overall performance (see Appendix B).
Finally, a series of simple logistic GLMs
revealed nine significant craniometric characteristics that can be used for determination of the European badger sexes (Table 2).
Results from the GLMs
for April, June, August, September and October, showed that temperature, salinity, dissolved oxygen, and Secchi disk depth were not significant predictor variables for catch of black sea bass at the 20 sampling sites (P>0.05).
There is little point in applying machine learning to small and simple datasets where a GLM
The GSEM combines the capabilities of structural equation models (SEMs) and generalized linear models (GLMs
scapularis nymphs, we ran univariate negative binomial generalized linear models (GLMs
.nb in the MASS package  in R ) for all land cover buffer sizes.