Mathematica 9 is now available
THIS IS DOCUMENTATION FOR AN OBSOLETE PRODUCT.
SEE THE DOCUMENTATION CENTER FOR THE LATEST INFORMATION.
Mathematica > Data Manipulation > Numerical Data > Statistical Model Analysis > LinkFunction >
Mathematica > Data Manipulation > Statistical Data Analysis > Statistical Model Analysis > LinkFunction >
Mathematica > Mathematics and Algorithms > Statistical Data Analysis > Statistical Model Analysis > LinkFunction >

LinkFunction

LinkFunction
is an option for GeneralizedLinearModelFit that specifies the link function for the generalized linear model.
  • The link function is an invertible function in the generalized linear model .
Automaticautomatically determined
"name"named link function
ginvertible function
  • The canonical link functions are as follows:
"LogitLink"used for "Binomial"
"ReciprocalLink"used for "Gamma"
"IdentityLink"used for "Gaussian"
"InverseSquareLink"used for "InverseGaussian"
"LogLink"used for "Poisson"
  • For models, is used by default.
  • Other common link functions for binomial data include:
"ProbitLink"
"CauchitLink"
"LogLogLink"
"LogComplementLink"
"ComplementaryLogLogLink"
"OddsPowerLink"
  • Other common link functions for count data include:
"NegativeBinomialLink"
  • Other common link functions for positive real-valued data include:
"PowerLink"
  • For , , and , the additional parameter can be given by LinkFunction. The parameter can be any real value for and and any positive value for .
  • With setting LinkFunction->g, g can be any pure function that is real-valued and invertible on the response domain for the model.
Fit a Poisson model with canonical Log link:
Use a named link:
Use a pure function for a shifted Sqrt link:
In[1]:=
Click for copyable input
Fit a Poisson model with canonical Log link:
In[2]:=
Click for copyable input
Out[2]=
Use a named link:
In[3]:=
Click for copyable input
Out[3]=
Use a pure function for a shifted Sqrt link:
In[4]:=
Click for copyable input
Out[4]=
New in 7
Ask a question about this page  |  Suggest an improvement  |  Leave a message for the team
Format:   HTML  |  CDF