This article relies largely or entirely on a single source. Relevant discussion may be found on the talk page. Please help improve this article by introducing citations to additional sources.Find sources: "Vector generalized linear model" – news · newspapers · books · scholar · JSTOR (June 2020)

Part of a series on |

Regression analysis |
---|

Models |

Estimation |

Background |

In statistics, the class of **vector generalized linear models** (**VGLMs**) was proposed to
enlarge the scope of models catered for by generalized linear models (**GLMs**).
In particular, VGLMs allow for response variables outside the classical exponential family
and for more than one parameter. Each parameter (not necessarily a mean) can be transformed by a *link function*.
The VGLM framework is also large enough to naturally accommodate multiple responses; these are
several independent responses each coming from a particular statistical distribution with
possibly different parameter values.

Vector generalized linear models are described in detail in Yee (2015).^{[1]}
The central algorithm adopted is the iteratively reweighted least squares method,
for maximum likelihood estimation of usually all the model parameters. In particular,
Fisher scoring is implemented by such, which, for most models,
uses the first and expected second derivatives of the log-likelihood function.

GLMs essentially cover one-parameter models from the classical exponential family, and include 3 of the most important statistical regression models: the linear model, Poisson regression for counts, and logistic regression for binary responses. However, the exponential family is far too limiting for regular data analysis. For example, for counts, zero-inflation, zero-truncation and overdispersion are regularly encountered, and the makeshift adaptations made to the binomial and Poisson models in the form of quasi-binomial and quasi-Poisson can be argued as being ad hoc and unsatisfactory. But the VGLM framework readily handles models such as zero-inflated Poisson regression, zero-altered Poisson (hurdle) regression, positive-Poisson regression, and negative binomial regression. As another example, for the linear model, the variance of a normal distribution is relegated as a scale parameter and it is treated often as a nuisance parameter (if it is considered as a parameter at all). But the VGLM framework allows the variance to be modelled using covariates.

As a whole, one can loosely think of VGLMs as GLMs that handle many models
outside the classical exponential family and are not restricted to estimating
a single mean.
During estimation,
rather than using weighted least squares
during IRLS, one uses generalized least squares to handle the
correlation between the *M* linear predictors.

We suppose that the response or outcome or the dependent variable(s), , are assumed to be generated from a particular distribution. Most distributions are univariate, so that , and an example of is the bivariate normal distribution.

Sometimes we write our data as
for . Each of the *n* observations are considered to be
independent.
Then .
The are known positive prior weights, and often .

The explanatory or independent variables are written ,
or when *i* is needed, as .
Usually there is an *intercept*, in which case
or .

Actually, the VGLM framework allows for *S* responses, each of dimension .
In the above *S* = 1. Hence the dimension of is more generally . One handles *S* responses by code such
as `vglm(cbind(y1, y2, y3) ~ x2 + x3, ..., data = mydata)`

for *S* = 3.
To simplify things, most of this article has *S* = 1.

The VGLM usually consists of four elements:

- 1. A probability density function or probability mass function from some statistical distribution which has a log-likelihood , first derivatives and expected information matrix that can be computed. The model is required to satisfy the usual MLE regularity conditions.
- 2. Linear predictors described below to model each parameter ,
- 3. Link functions such that
- 4.
*Constraint matrices*for each of full column-rank and known.

Each *linear predictor* is a quantity which incorporates
information about the independent variables into the model.
The symbol (Greek "eta")
denotes a linear predictor and a subscript *j* is used to denote the *j*th one.
It relates the *j*th parameter to the explanatory variables, and
is expressed as linear combinations (thus, "linear")
of unknown parameters
i.e., of regression coefficients .

The *j*th parameter, , of the distribution depends on the
independent variables, through

Let be the vector of
all the linear predictors. (For convenience we always let
be of dimension *M*).
Thus *all* the covariates comprising potentially affect *all* the parameters through the linear predictors . Later, we will allow the linear predictors to be generalized to additive predictors, which is the sum of smooth functions of each and each function is estimated from the data.

Each link function provides the relationship between a linear predictor and a
parameter of the distribution.
There are many commonly used link functions, and their choice can be somewhat arbitrary. It makes sense to try to match the domain of the link function to
the range of the distribution's parameter value.
Notice above that the allows a different link function for each parameter.
They have similar properties as with generalized linear models, for example,
common link functions include the *logit* link for parameters in ,
and the log link for positive parameters. The `VGAM`

package has function `identitylink()`

for parameters that can assume both positive and negative values.

More generally, the VGLM framework allows for any linear constraints between the regression coefficients of each linear predictors. For example, we may want to set some to be equal to 0, or constraint some of them to be equal. We have

where the are the *constraint matrices*.
Each constraint matrix is known and prespecified, and has *M* rows, and between 1 and *M* columns. The elements of constraint matrices are finite-valued, and often they are just 0 or 1.
For example, the value 0 effectively omits that element while a 1 includes it.
It is common for some models to have a *parallelism* assumption, which means that
for , and
for some models, for too.
The special case when for
all is known as *trivial constraints*; all the
regression coefficients are estimated and are unrelated.
And is known as an *intercept-only* parameter
if the *j*th row of all the are equal to for , i.e., equals an intercept only. Intercept-only parameters are thus modelled as simply as possible, as a scalar.

The unknown parameters, , are typically estimated by the method of maximum likelihood. All the regression coefficients may be put into a matrix as follows:

With even more generally, one can allow the value of a variable
to have a different value for each .
For example, if each linear predictor is for a different time point then
one might have a time-varying covariate.
For example,
in discrete choice models, one has
*conditional* logit models,
*nested* logit models,
*generalized* logit models,
and the like, to distinguish between certain variants and
fit a multinomial logit model to, e.g., transport choices.
A variable such as cost differs depending on the choice, for example,
taxi is more expensive than bus, which is more expensive than walking.
The `xij`

facility of `VGAM`

allows one to
generalize
to .

The most general formula is

Here the is an optional *offset*; which translates
to be a matrix in practice.
The `VGAM`

package has an `xij`

argument that allows
the successive elements of the diagonal matrix to be inputted.

Yee (2015)^{[1]} describes an R package
implementation in the
called VGAM.^{[2]}
Currently this software fits approximately 150 models/distributions.
The central modelling functions are `vglm()`

and `vgam()`

.
The `family`

argument is assigned a *VGAM family function*,
e.g., `family = negbinomial`

for negative binomial regression,
`family = poissonff`

for Poisson regression,
`family = propodds`

for the *proportional odd model* or
*cumulative logit model* for ordinal categorical regression.

We are maximizing a log-likelihood

where the are positive and known *prior weights*.
The maximum likelihood estimates can be found
using an iteratively reweighted least squares algorithm using
Fisher's scoring method, with updates of the form:

where is
the Fisher information matrix at iteration *a*.
It is also called the *expected information matrix*, or *EIM*.

For the computation, the (small) *model matrix* constructed
from the RHS of the formula in `vglm()`

and the constraint matrices are combined to form a *big* model matrix.
The IRLS is applied to this big **X**. This matrix is known as the VLM
matrix, since the *vector linear model* is the underlying least squares
problem being solved. A VLM is a weighted multivariate regression where the
variance-covariance matrix for each row of the response matrix is not
necessarily the same, and is known.
(In classical multivariate regression, all the errors have the
same variance-covariance matrix, and it is unknown).
In particular, the VLM minimizes the weighted sum of squares

This quantity is minimized at each IRLS iteration.
The *working responses* (also known as *pseudo-response* and *adjusted dependent vectors*) are

where the are known as *working weights* or *working weight matrices*. They are symmetric and positive-definite. Using the EIM helps ensure that they are all positive-definite (and not just the sum of them) over much of the parameter space. In contrast, using Newton–Raphson would mean the observed information matrices would be used, and these tend to be positive-definite in a smaller subset of the parameter space.

Computationally, the Cholesky decomposition is used to invert the working weight matrices and to convert the overall generalized least squares problem into an ordinary least squares problem.

Of course, all generalized linear models are a special cases of VGLMs. But we often estimate all parameters by full maximum likelihood estimation rather than using the method of moments for the scale parameter.

If the response variable is an ordinal measurement with *M* + 1 *levels*, then one may fit a model function of the form:

- where

for
Different links *g* lead to proportional odds models or ordered probit models,
e.g., the `VGAM`

family function `cumulative(link = probit)`

assigns a probit link to the cumulative
probabilities, therefore this model is also called the *cumulative probit model*.
In general they are called *cumulative link models*.

For categorical and multinomial distributions, the fitted values are an (*M* + 1)-vector of probabilities, with the property that all probabilities add up to 1. Each probability indicates the likelihood of occurrence of one of the *M* + 1 possible values.

If the response variable is a nominal measurement, or the data do not satisfy the assumptions of an ordered model, then one may fit a model of the following form:

for The above link is sometimes called the *multilogit* link,
and the model is called the multinomial logit model.
It is common to choose the first or the last level of the response as the
*reference* or *baseline* group; the above uses the last level.
The `VGAM`

family function `multinomial()`

fits the above model,
and it has an argument called `refLevel`

that can be assigned
the level used for as the reference group.

Classical GLM theory performs Poisson regression for count data.
The link is typically the logarithm, which is known as the *canonical link*.
The variance function is proportional to the mean:

where the dispersion parameter is typically fixed at exactly one. When it is not, the resulting quasi-likelihood model is often described as Poisson with overdispersion, or *quasi-Poisson*; then is commonly estimated by the method-of-moments and as such,
confidence intervals for are difficult to obtain.

In contrast, VGLMs offer a much richer set of models to handle overdispersion with respect to the Poisson, e.g., the negative binomial distribution and several variants thereof. Another count regression model is the *generalized Poisson distribution*. Other possible models are the *zeta distribution* and the *Zipf distribution*.

RR-VGLMs are VGLMs where a subset of
the **B** matrix is of a lower rank.
Without loss of generality, suppose that is a partition of the covariate vector. Then the part of the **B** matrix corresponding to is of the form where and
are thin matrices (i.e., with *R* columns), e.g., vectors if the rank *R* = 1. RR-VGLMs potentially offer several advantages when applied to certain
models and data sets. Firstly, if *M* and *p* are large then the number of regression coefficients
that are estimated by VGLMs is large (). Then RR-VGLMs can reduce the number of estimated regression coefficients enormously if *R* is low, e.g., *R* = 1
or *R* = 2. An example of a model where this is particularly useful is the RR-multinomial logit model, also known as the *stereotype model*.
Secondly,
is an *R*-vector of latent variables, and often these can be usefully interpreted.
If *R* = 1 then we can write
so that the latent variable comprises loadings on the explanatory variables.
It may be seen that RR-VGLMs take optimal linear combinations of the
and then a VGLM is fitted to the explanatory variables . Thirdly, a biplot can be produced if *R* = 2 , and this allows the model to be visualized.

It can be shown that RR-VGLMs are simply VGLMs where the constraint matrices for
the variables in are unknown and to be estimated.
It then transpires that for
such variables.
RR-VGLMs can be estimated by an *alternating* algorithm which fixes
and estimates and then fixes and estimates , etc.

In practice, some uniqueness constraints are needed for
and/or . In `VGAM`

, the `rrvglm()`

function uses *corner constraints* by default, which means that the top *R* rows of is set to . RR-VGLMs were proposed in 2003.^{[3]}

A special case of RR-VGLMs is when *R* = 1 and *M* = 2. This is *dimension reduction* from 2 parameters to 1 parameter. Then it can be shown that

where elements and are estimated. Equivalently,

This formula provides a coupling of and . It induces a relationship between two parameters of a model that can be useful, e.g., for modelling a mean-variance relationship. Sometimes there is some choice of link functions, therefore it offers a little flexibility when coupling the two parameters, e.g., a logit, probit, cauchit or cloglog link for parameters in the unit interval. The above formula is particularly useful for the negative binomial distribution, so that the RR-NB has variance function

This has been called the *NB-P* variant by some authors. The and are estimated, and it is also possible to obtain approximate confidence intervals for them too.

Incidentally, several other useful NB variants can also be fitted, with the help of selecting the right combination of constraint matrices. For example, *NB* − 1, *NB* − 2 (`negbinomial()`

default), *NB* − *H*; see Yee (2014)^{[4]} and Table 11.3 of Yee (2015).^{[1]}

The subclass of *row-column interaction models*
(RCIMs) has also been proposed; these are a special type of RR-VGLM.
RCIMs apply only to a matrix **Y** response and there are
no explicit explanatory variables .
Instead, indicator variables for each row and column are explicitly set up, and an order-*R*
interaction of the form is allowed.
Special cases of this type of model include the *Goodman RC association model*
and the quasi-variances methodology as implemented by the `qvcalc`

R package.

RCIMs can be defined as a RR-VGLM applied to **Y** with

For the Goodman RC association model, we have so that
if *R* = 0 then it is a Poisson regression fitted to a matrix of counts with row effects and column effects; this has a similar idea to a no-interaction two-way ANOVA model.

Another example of a RCIM is if is the identity link and the parameter is the median and the model corresponds to an asymmetric Laplace distribution; then a no-interaction RCIM is similar to a technique called *median polish*.

In `VGAM`

, `rcim()`

and `grc()`

functions fit the above models.
And also Yee and Hadi (2014)^{[5]}
show that RCIMs can be used to fit unconstrained quadratic ordination
models to species data; this is an example of indirect gradient analysis in
ordination (a topic in statistical ecology).

*Vector generalized additive models* (VGAMs) are a major
extension to VGLMs in which the linear predictor is not restricted to be
linear in the covariates but is the
sum of smoothing functions applied to the :

where
These are *M* *additive predictors*.
Each smooth function is estimated from the data.
Thus VGLMs are *model-driven* while VGAMs are *data-driven*.
Currently, only smoothing splines are implemented in the `VGAM`

package.
For *M* > 1 they are actually *vector splines*, which estimate the component functions
in simultaneously.
Of course, one could use regression splines with VGLMs.
The motivation behind VGAMs is similar to
that of
Hastie and Tibshirani (1990)^{[6]}
and
Wood (2017).^{[7]}
VGAMs were proposed in 1996
.^{[8]}

Currently, work is being done to estimate VGAMs using *P-splines*
of Eilers and Marx (1996)
.^{[9]}
This allows for several advantages over using smoothing splines and vector backfitting, such as the
ability to perform automatic smoothing parameter selection easier.

These add on a quadratic in the latent variable to the RR-VGLM class.
The result is a bell-shaped curve can be fitted to each response, as
a function of the latent variable.
For *R* = 2, one has bell-shaped surfaces as a function of the 2
latent variables---somewhat similar to a
bivariate normal distribution.
Particular applications of QRR-VGLMs can be found in ecology,
in a field of multivariate analysis called ordination.

As a specific rank-1 example of a QRR-VGLM,
consider Poisson data with *S* species.
The model for Species *s* is the Poisson regression

for . The right-most parameterization which uses the symbols has particular ecological meaning, because they relate to the species *abundance*, *optimum* and *tolerance* respectively. For example, the tolerance is a measure of niche width, and a large value means that that species can live in a wide range of environments. In the above equation, one would need in order
to obtain a bell-shaped curve.

QRR-VGLMs fit Gaussian ordination models by maximum likelihood estimation, and
they are an example of direct gradient analysis.
The `cqo()`

function in the `VGAM`

package currently
calls `optim()`

to search for the optimal
, and given that, it is easy to calculate
the site scores and fit a suitable generalized linear model to that.
The function is named after the acronym CQO, which stands for
*constrained quadratic ordination*: the *constrained* is for direct
gradient analysis (there are environmental variables, and a linear combination
of these is taken as the latent variable) and the *quadratic* is for the
quadratic form in the latent variables
on the scale.
Unfortunately QRR-VGLMs are sensitive to outliers in both the response
and explanatory variables, as well as being computationally expensive, and
may give a local solution rather than a global solution.
QRR-VGLMs were proposed in 2004.^{[10]}