R devel - Aug 2000 - predict.lm is broken in 1.1.0-patched (2000-August-7) (PR#626)

predict.lm has been broken by recent changes to the patched branch.

It fails for all singular fits.  An example:

library(MASS)
data(quine)
quine.hi <- aov(log(Days + 2.5) ~ .^4, quine)
quine.nxt <- update(quine.hi, . ~ . - Eth:Sex:Age:Lrn)
predict(quine.nxt)
  1   2   3   4   5   6   7   8   9  10  11  12  13  14  15  16  17  18  19  20 
 NA  NA  NA  NA  NA  NA  NA  NA  NA  NA  NA  NA  NA  NA  NA  NA  NA  NA  NA  NA 
 21  22  23  24  25  26  27  28  29  30  31  32  33  34  35  36  37  38  39  40 
 NA  NA  NA  NA  NA  NA  NA  NA  NA  NA  NA  NA  NA  NA  NA  NA  NA  NA  NA  NA 
 41  42  43  44  45  46  47  48  49  50  51  52  53  54  55  56  57  58  59  60 
 NA  NA  NA  NA  NA  NA  NA  NA  NA  NA  NA  NA  NA  NA  NA  NA  NA  NA  NA  NA 
 61  62  63  64  65  66  67  68  69  70  71  72  73  74  75  76  77  78  79  80 
 NA  NA  NA  NA  NA  NA  NA  NA  NA  NA  NA  NA  NA  NA  NA  NA  NA  NA  NA  NA 
 81  82  83  84  85  86  87  88  89  90  91  92  93  94  95  96  97  98  99 100 
 NA  NA  NA  NA  NA  NA  NA  NA  NA  NA  NA  NA  NA  NA  NA  NA  NA  NA  NA  NA 
101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 
 NA  NA  NA  NA  NA  NA  NA  NA  NA  NA  NA  NA  NA  NA  NA  NA  NA  NA  NA  NA 
121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 
 NA  NA  NA  NA  NA  NA  NA  NA  NA  NA  NA  NA  NA  NA  NA  NA  NA  NA  NA  NA 
141 142 143 144 145 146 
 NA  NA  NA  NA  NA  NA 

The answeer should be (e.g. R 1.1.0)
> predict(quine.nxt)
      1       2       3       4       5       6       7       8 
2.19247 2.19247 2.19247 2.68311 2.68311 2.68311 2.68311 2.68311 
      9      10      11      12      13      14      15      16 
2.37260 2.37260 2.37260 2.53960 2.53960 3.53516 3.53516 3.53516 
     17      18      19      20      21      22      23      24 
3.53516 3.24298 3.24298 3.24298 3.24298 3.24298 3.24298 3.24298 
     25      26      27      28      29      30      31      32 
3.31368 3.31368 3.31368 3.31368 3.31368 3.31368 3.31368 2.03746 
...

I believe this is a lot worse than that which the changes were intended to
fix. (Are we not supposed to be ultra-cautious about not making things
worse in that branch?)

--please do not edit the information below--

Version:
 platform = sparc-sun-solaris2.6
 arch = sparc
 os = solaris2.6
 system = sparc, solaris2.6
status   Patched             
major    1                   
minor    1.0                 
year     2000                
month    August              
day      7                   
language R                   

Search Path:
 .GlobalEnv, Autoloads, package:base

-- 
Brian D. Ripley,                  ripley@stats.ox.ac.uk
Professor of Applied Statistics,  http://www.stats.ox.ac.uk/~ripley/
University of Oxford,             Tel:  +44 1865 272861 (self)
1 South Parks Road,                     +44 1865 272860 (secr)
Oxford OX1 3TG, UK                Fax:  +44 1865 272595


-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-
r-devel mailing list -- Read http://www.ci.tuwien.ac.at/~hornik/R/R-FAQ.html
Send "info", "help", or "[un]subscribe"
(in the "body", not the subject !)  To:
r-devel-request@stat.math.ethz.ch
_._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._

ripley@stats.ox.ac.uk writes:
> predict.lm has been broken by recent changes to the patched branch.
> 
> It fails for all singular fits.  An example:
> 
> library(MASS)
> data(quine)
> quine.hi <- aov(log(Days + 2.5) ~ .^4, quine)
> quine.nxt <- update(quine.hi, . ~ . - Eth:Sex:Age:Lrn)
> predict(quine.nxt)
>   1   2   3   4   5   6   7   8   9  10  11  12  13  14  15  16  17  18  19
20
>  NA  NA  NA  NA  NA  NA  NA  NA  NA  NA  NA  NA  NA  NA  NA  NA  NA  NA  NA
NA
...
> I believe this is a lot worse than that which the changes were intended to
> fix. (Are we not supposed to be ultra-cautious about not making things
> worse in that branch?)
Sometimes the fix for one bug uncovers another. We have
> coef.aov
function (object, ...) 
{
    z <- object$coef
    z[!is.na(z)]
}
> coef.lm 
function (object, ...) 
object$coefficients

and both of these cannot be right when predict.lm calls coef(object)
and object$coefficients contains NA's. Using object$coefficients
inside predict.lm fixes things, but I'm still not happy with the two
different conventions for coef...

-- 
   O__  ---- Peter Dalgaard             Blegdamsvej 3  
  c/ /'_ --- Dept. of Biostatistics     2200 Cph. N   
 (*) \(*) -- University of Copenhagen   Denmark      Ph: (+45) 35327918
~~~~~~~~~~ - (p.dalgaard@biostat.ku.dk)             FAX: (+45) 35327907
-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-
r-devel mailing list -- Read http://www.ci.tuwien.ac.at/~hornik/R/R-FAQ.html
Send "info", "help", or "[un]subscribe"
(in the "body", not the subject !)  To:
r-devel-request@stat.math.ethz.ch
_._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._