corstr="exchangeable", intercept=T, summarize=F)

#

Modelling output directed to an object called gee.mod.exch (an arbitrary name)

#

x = explanatory variable,

#

y = response variable,

#

id=cluster (individual) identifier,

#

varfun = distribution of response variable (Normal),

#

link = link function (identity),

#

corstr = correlation structure (exchangeable),

#

intercept=T fits a regression intercept

#

summarizesF allows output to be saved as an Splus object (i.e. gee.mod.exch)

#

To interpret output:

#

Robust and model-based standard errors can only be obtained from the square

#

root of the diagonal elements of the robust and model-based variance-covariance

#

matrices called "gee.mod.exch$robvar" and "gee.mod.exch$naiwar"

#

respectively. Regression coefficients are held in the vector "gee.mod.exch$regest".

#

The intra-class correlation coefficient may be obtained from the off-diagonal

#

elements of the working correlation matrix called "gee.mod.exch$worcor".

# denotes a comment in Splus. Unbolded text represents active Splus code.

Figure 3. S-plus code required to fit an GEE model with an exchangeable correlation structure

# denotes a comment in Splus. Unbolded text represents active Splus code.

Figure 3. S-plus code required to fit an GEE model with an exchangeable correlation structure

Table II. A GEE model with an exchangeable correlation structure

Parameter |
Value |

Fixed regression coefficients |
Was this article helpful? |

## Post a comment