R help - Sep 2010 - drop.terms and [.terms ignores intercept

Dear R-help list

The functions drop.terms and [.terms ignores if the intercept
has been explicitly removed. Is that a deliberate feature?

For instance,

drop.terms(terms(~a+b-1),1)
terms(~a+b-1)[2]

R version 2.12.0 Under development (unstable) (2010-09-13 r52905)

Best, Niels

-- 
Niels Richard Hansen                     Web:   www.math.ku.dk/~richard	
Associate Professor                      Email: Niels.R.Hansen at math.ku.dk
Department of Mathematical Sciences             nielsrichardhansen at gmail.com
University of Copenhagen                 Skype: nielsrichardhansen.dk	
Universitetsparken 5                     Phone: +45 353 20783 (office)	
2100 Copenhagen ?                               +45 2859 0765 (mobile)
Denmark

On Sep 29, 2010, at 10:29 , Niels Richard Hansen wrote:
> The functions drop.terms and [.terms ignores if the intercept
> has been explicitly removed. Is that a deliberate feature?
Perhaps rather an unimplemented one. The root cause is that both functions use
reformulate() on the "term.labels" attribute, and there is no way to
specify that you want to reformulate into a no-intercept formula. On the other
hand, the modeling code will happily proceed with a no-intercept model even if
there is no "-1" in formula part of a terms object, e.g.
> x <- terms(y~a+b)
> attr(x,"intercept") <- 0
> lm(x)
Call:
lm(formula = x)

Coefficients:
     a       b  
0.2263  0.4178  
> formula(x)
y ~ a + b

so I suppose that there is no really good excuse not to carry the
"intercept" attribute over. As usual, with code as old as this, there
is always the risk that something actually relies on current behavior.


-- 
Peter Dalgaard
Center for Statistics, Copenhagen Business School
Solbjerg Plads 3, 2000 Frederiksberg, Denmark
Phone: (+45)38153501
Email: pd.mes at cbs.dk  Priv: PDalgd at gmail.com

On 29/09/10 11.13, peter dalgaard wrote:
>
> On Sep 29, 2010, at 10:29 , Niels Richard Hansen wrote:
>
>> The functions drop.terms and [.terms ignores if the intercept has been
>> explicitly removed. Is that a deliberate feature?
>
> Perhaps rather an unimplemented one. The root cause is that both functions
> use reformulate() on the "term.labels" attribute, and there is no
way to
> specify that you want to reformulate into a no-intercept formula. On the
> other hand, the modeling code will happily proceed with a no-intercept
model
> even if there is no "-1" in formula part of a terms object, e.g.
>
>> x<- terms(y~a+b) attr(x,"intercept")<- 0 lm(x)
>
> Call: lm(formula = x)
>
> Coefficients: a       b 0.2263  0.4178
>
>> formula(x)
> y ~ a + b
>
> so I suppose that there is no really good excuse not to carry the
"intercept"
> attribute over. As usual, with code as old as this, there is always the
risk
> that something actually relies on current behavior.
Thanks Peter. I expected something like that.

As I see it, the intercept (or more often, removal of the intercept)
is treated differently from other terms, and information on the presence
of an intercept is stored in an attribute of the terms object rather
than the formula. This attribute is not copied when using drop.terms
or [.terms. I believe it should be, but as you say Peter, some may rely
on the way the functions behave now.

So basically, you should use code like

x <- terms(y~a+b-1)
z <- x[2]
attr(z, "intercept") <-  attr(x, "intercept")

to make sure that the intercept information is preserved.

- Niels
>
>
-- 
Niels Richard Hansen                     Web:   www.math.ku.dk/~richard	
Associate Professor                      Email: Niels.R.Hansen at math.ku.dk
Department of Mathematical Sciences             nielsrichardhansen at gmail.com
University of Copenhagen                 Skype: nielsrichardhansen.dk	
Universitetsparken 5                     Phone: +45 353 20783 (office)	
2100 Copenhagen ?                               +45 2859 0765 (mobile)
Denmark