R help - Apr 2004 - predict.gl( ..., type="terms" )

When I do:
> apc <- glm( D ~ ns( Ax, knots=seq(50,80,10), Bo=c(40,90) ) +
+                 ns( Cx, knots=seq(1880,1940,20), Bo=c(1840,1960) ) +
+                 ns( Px, knots=seq(1960,1980,10), Bo=c(1940,2000) ) +
+                 offset( log( Y ) ),
+             family=poisson )
> pterm <-  predict( apc, type="terms" )
> plink <-  predict( apc, type="link" )
> ( apply( pterm, 1, sum ) + log( Y ) - plink )[1:10]
      1       2       3       4       5       6       7       8       9
10 
6.85047 6.85047 6.85047 6.85047 6.85047 6.85047 6.85047 6.85047 6.85047
6.85047 
> coef( apc )[1] 
(Intercept) 
  -13.61998 
>From the help page for predict.glm I would have expected that the
constant 6.85 
was -intercept.

What am I missing from predict.glm? (or from splines?)

Bendix Carstensen

----------------------
Bendix Carstensen
Senior Statistician
Steno Diabetes Center
Niels Steensens Vej 2
DK-2820 Gentofte
Denmark
tel: +45 44 43 87 38
mob: +45 30 75 87 38
fax: +45 44 43 07 06
bxc at steno.dk
www.biostat.ku.dk/~bxc

"BXC (Bendix Carstensen)" <bxc at steno.dk> writes:
> > apc <- glm( D ~ ns( Ax, knots=seq(50,80,10), Bo=c(40,90) ) +
> +                 ns( Cx, knots=seq(1880,1940,20), Bo=c(1840,1960) ) +
> +                 ns( Px, knots=seq(1960,1980,10), Bo=c(1940,2000) ) +
> +                 offset( log( Y ) ),
> +             family=poisson )
> > pterm <-  predict( apc, type="terms" )
> > plink <-  predict( apc, type="link" )
> > ( apply( pterm, 1, sum ) + log( Y ) - plink )[1:10]
>       1       2       3       4       5       6       7       8       9
> 10 
> 6.85047 6.85047 6.85047 6.85047 6.85047 6.85047 6.85047 6.85047 6.85047
> 6.85047 
> > coef( apc )[1] 
> (Intercept) 
>   -13.61998 
> 
> >From the help page for predict.glm I would have expected that the
> constant 6.85 
> was -intercept.
> 
> What am I missing from predict.glm? (or from splines?)
Same thing that you're missing from predict(..., type="terms") in
general.
Try

y <- rnorm(10)
x <- runif(10)
apc <- lm(y~x)
predict(apc,type="terms")
predict(apc,type="terms") - predict(apc,type="response")
mean(predict(apc,type="terms"))

and I think enlightenment will follow.

-- 
   O__  ---- Peter Dalgaard             Blegdamsvej 3  
  c/ /'_ --- Dept. of Biostatistics     2200 Cph. N   
 (*) \(*) -- University of Copenhagen   Denmark      Ph: (+45) 35327918
~~~~~~~~~~ - (p.dalgaard at biostat.ku.dk)             FAX: (+45) 35327907