adding regression curve to xyplot

Dear R-listers,

It seems that predict() behaves differently within panel.xyplot.  Am I
doing something stupid?

Thanks,

Carlisle

First, without xyplot():
lmtest <- lm(t~s,data=subset(P100,whichLon100==1 & whichLat100==1))
lmtest
Call:
lm(formula = s ~ t, data = subset(P100, whichLon100 == 1 & whichLat100 ==
   1))

Coefficients:
(Intercept)            t
    33.3307       0.1393
range(P100$t)
[1]  4.050469 24.514543
predict(lmtest,newdata=data.frame(t=range(P100$t)))
1        2
33.89501 36.74620

As expected, predict gives two values.  But inside xyplot() predict gives
300 values:
xyplot(t~s|factor(lonLabels[whichLon100])*factor(latLabels[whichLat100]),
+   data=P100,pch=".",
+   panel=function(x,y,...){panel.xyplot(x,y,...)
+                           thislm <- lm(x~y)
+                      print(thislm)
+                           newt <- range(P100$t)
+                      print(newt)
+                           news <-
as.vector(predict(thislm,newdata=data.frame(t=newt)))
+                      print(news)
+                           llines(news,newt,col="red")})

Call:
lm(formula = x ~ y)

Coefficients:
(Intercept)            y
    33.3307       0.1393

[1]  4.050469 24.514543
  [1] 34.16173 35.31284 34.95317 35.62314 34.64448 34.06225 34.19688 34.03391
  [9] 33.90860 35.28849 35.29911 34.61618 35.19837 35.28966 34.91978 35.32388
 [17] 35.39351 35.54533 35.58760 34.77039 35.07226 35.35643 35.40594 33.98609
 [25] 33.96461 33.91158 34.12389 34.33804 34.72235 34.63541 36.20420 36.13635
 [33] 34.60623 34.32401 35.59768 34.68576 34.64661 34.58804 34.74548 35.09062
 [41] 34.70290 34.22428 35.12743 34.90726 34.52325 34.90041 35.46984 34.79572
 [49] 35.47026 35.35481 34.70722 34.08883 33.97586 33.94797 33.96165 33.97715
 [57] 33.94886 33.99715 34.10887 34.72597 34.20534 34.45016 35.00769 34.56742
 [65] 34.76119 34.87190 34.60760 34.50113 34.56594 34.65954 34.57076 34.91072
 [73] 34.97322 34.99036 34.48229 34.58436 34.56767 34.94204 34.95929 34.89107
 [81] 34.98846 34.51328 34.74252 34.61337 34.62212 34.70953 34.74080 34.71630
 [89] 34.68599 35.58629 34.77031 34.49366 34.49873 34.48112 35.11327 34.62598
 [97] 34.50436 34.50705 34.70170 35.01278 35.12273 35.15541 35.06295 35.15366
[105] 35.15089 35.11312 35.09834 35.05271 34.84838 34.52225 34.34650 34.32637
[113] 34.42911 34.79282 35.06830 35.11559 35.25736 34.43281 34.45208 34.58407
[121] 35.28598 34.34246 34.32532 34.47483 34.79309 34.02772 34.01553 35.33372
[129] 34.19273 34.07622 34.52820 35.27539 35.37703 34.70560 34.77501 34.52084
[137] 34.68773 35.19217 35.69538 34.21886 34.17226 34.94784 34.03079 33.95669
[145] 33.92002 33.92972 33.96568 33.92996 33.94236 33.93887 33.92059 33.91746
[153] 33.98369 34.17190 34.05327 33.96500 34.16136 33.98172 34.02463 34.00802
[161] 34.05238 34.05088 34.16828 34.14888 33.94711 33.96567 34.12450 33.98988
[169] 33.98670 33.98489 34.05447 34.03689 34.17137 34.31777 34.05027 33.97821
[177] 33.99281 34.00209 34.07560 33.98840 33.98930 33.98548 33.99127 35.23652
[185] 35.17034 35.53546 35.53861 34.02384 34.03402 33.93877 33.93540 33.90636
[193] 33.92267 34.57678 34.49457 34.46038 34.64318 35.06561 35.06680 34.76466
[201] 34.00418 35.48020 35.12687 35.28180 35.07281 35.23499 35.14693 34.07914
[209] 33.99047 34.00648 34.10073 34.00784 33.99185 33.98191 34.03089 33.96755
[217] 34.17042 34.10803 34.12671 34.09942 34.25003 34.06542 34.07293 34.19048
[225] 34.06594 34.01929 34.00155 34.11555 33.89754 33.89527 33.89501 33.93433
[233] 33.89881 33.94577 33.96603 33.95199 34.00284 34.12974 33.98116 34.08882
[241] 34.00953 34.98977 34.72824 34.68920 35.15207 34.59716 34.64580 34.25087
[249] 34.96008 34.62607 35.32346 35.16993 34.46759 34.47157 34.46978 34.46092
[257] 34.46561 34.46489 34.49970 34.49599 34.55441 34.52048 34.58480 34.59267
[265] 34.50082 34.57311 34.60421 34.57043 34.53131 34.46365 34.41946 34.74566
[273] 34.64593 34.67921 34.65706 34.71625 34.61487 34.61687 34.58899 34.47081
[281] 34.58366 35.66823 35.61883 34.71477 34.72518 34.69732 34.66265 34.65186
[289] 34.81468 34.74725 34.77658 34.77191 34.81481 34.76427 34.70751 34.96763
[297] 35.11322 35.16852 34.94393 34.89605
Error in xy.coords(x, y) : x and y lengths differ

Dear R-listers,

It seems that predict() behaves differently within panel.xyplot.  Am I
doing something stupid?
You fitted the model lm(x ~ y) and supplied new values for t, not y.

Using panel.abline would be a bit easier: just call panel.abline(thislm) 
from your panel function.
Thanks,

Carlisle

First, without xyplot():
lmtest <- lm(t~s,data=subset(P100,whichLon100==1 & whichLat100==1))
lmtest
Call:
lm(formula = s ~ t, data = subset(P100, whichLon100 == 1 & whichLat100 ==
  1))

Coefficients:
(Intercept)            t
   33.3307       0.1393
range(P100$t)
[1]  4.050469 24.514543
predict(lmtest,newdata=data.frame(t=range(P100$t)))
      1        2
33.89501 36.74620

As expected, predict gives two values.  But inside xyplot() predict gives
300 values:

xyplot(t~s|factor(lonLabels[whichLon100])*factor(latLabels[whichLat100]),
+   data=P100,pch=".",
+   panel=function(x,y,...){panel.xyplot(x,y,...)
+                           thislm <- lm(x~y)
+                      print(thislm)
+                           newt <- range(P100$t)
+                      print(newt)
+                           news <-
as.vector(predict(thislm,newdata=data.frame(t=newt)))
+                      print(news)
+                           llines(news,newt,col="red")})

Call:
lm(formula = x ~ y)

Coefficients:
(Intercept)            y
   33.3307       0.1393

[1]  4.050469 24.514543
 [1] 34.16173 35.31284 34.95317 35.62314 34.64448 34.06225 34.19688 34.03391
 [9] 33.90860 35.28849 35.29911 34.61618 35.19837 35.28966 34.91978 35.32388
[17] 35.39351 35.54533 35.58760 34.77039 35.07226 35.35643 35.40594 33.98609
[25] 33.96461 33.91158 34.12389 34.33804 34.72235 34.63541 36.20420 36.13635
[33] 34.60623 34.32401 35.59768 34.68576 34.64661 34.58804 34.74548 35.09062
[41] 34.70290 34.22428 35.12743 34.90726 34.52325 34.90041 35.46984 34.79572
[49] 35.47026 35.35481 34.70722 34.08883 33.97586 33.94797 33.96165 33.97715
[57] 33.94886 33.99715 34.10887 34.72597 34.20534 34.45016 35.00769 34.56742
[65] 34.76119 34.87190 34.60760 34.50113 34.56594 34.65954 34.57076 34.91072
[73] 34.97322 34.99036 34.48229 34.58436 34.56767 34.94204 34.95929 34.89107
[81] 34.98846 34.51328 34.74252 34.61337 34.62212 34.70953 34.74080 34.71630
[89] 34.68599 35.58629 34.77031 34.49366 34.49873 34.48112 35.11327 34.62598
[97] 34.50436 34.50705 34.70170 35.01278 35.12273 35.15541 35.06295 35.15366
[105] 35.15089 35.11312 35.09834 35.05271 34.84838 34.52225 34.34650 34.32637
[113] 34.42911 34.79282 35.06830 35.11559 35.25736 34.43281 34.45208 34.58407
[121] 35.28598 34.34246 34.32532 34.47483 34.79309 34.02772 34.01553 35.33372
[129] 34.19273 34.07622 34.52820 35.27539 35.37703 34.70560 34.77501 34.52084
[137] 34.68773 35.19217 35.69538 34.21886 34.17226 34.94784 34.03079 33.95669
[145] 33.92002 33.92972 33.96568 33.92996 33.94236 33.93887 33.92059 33.91746
[153] 33.98369 34.17190 34.05327 33.96500 34.16136 33.98172 34.02463 34.00802
[161] 34.05238 34.05088 34.16828 34.14888 33.94711 33.96567 34.12450 33.98988
[169] 33.98670 33.98489 34.05447 34.03689 34.17137 34.31777 34.05027 33.97821
[177] 33.99281 34.00209 34.07560 33.98840 33.98930 33.98548 33.99127 35.23652
[185] 35.17034 35.53546 35.53861 34.02384 34.03402 33.93877 33.93540 33.90636
[193] 33.92267 34.57678 34.49457 34.46038 34.64318 35.06561 35.06680 34.76466
[201] 34.00418 35.48020 35.12687 35.28180 35.07281 35.23499 35.14693 34.07914
[209] 33.99047 34.00648 34.10073 34.00784 33.99185 33.98191 34.03089 33.96755
[217] 34.17042 34.10803 34.12671 34.09942 34.25003 34.06542 34.07293 34.19048
[225] 34.06594 34.01929 34.00155 34.11555 33.89754 33.89527 33.89501 33.93433
[233] 33.89881 33.94577 33.96603 33.95199 34.00284 34.12974 33.98116 34.08882
[241] 34.00953 34.98977 34.72824 34.68920 35.15207 34.59716 34.64580 34.25087
[249] 34.96008 34.62607 35.32346 35.16993 34.46759 34.47157 34.46978 34.46092
[257] 34.46561 34.46489 34.49970 34.49599 34.55441 34.52048 34.58480 34.59267
[265] 34.50082 34.57311 34.60421 34.57043 34.53131 34.46365 34.41946 34.74566
[273] 34.64593 34.67921 34.65706 34.71625 34.61487 34.61687 34.58899 34.47081
[281] 34.58366 35.66823 35.61883 34.71477 34.72518 34.69732 34.66265 34.65186
[289] 34.81468 34.74725 34.77658 34.77191 34.81481 34.76427 34.70751 34.96763
[297] 35.11322 35.16852 34.94393 34.89605
Error in xy.coords(x, y) : x and y lengths differ

______________________________________________
R-help at stat.math.ethz.ch mailing list
https://stat.ethz.ch/mailman/listinfo/r-help
PLEASE do read the posting guide! http://www.R-project.org/posting-guide.html

Brian D. Ripley,                  ripley at 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 272866 (PA)
Oxford OX1 3TG, UK                Fax:  +44 1865 272595
Dear R-listers,

It seems that predict() behaves differently within panel.xyplot.  Am
I doing something stupid?

Thanks,

Carlisle

First, without xyplot():
lmtest <- lm(t~s,data=subset(P100,whichLon100==1 & whichLat100==1))
lmtest
Call:
lm(formula = s ~ t, data = subset(P100, whichLon100 == 1 &
whichLat100 == 1))

Coefficients:
(Intercept)            t
    33.3307       0.1393

range(P100$t)
[1]  4.050469 24.514543

predict(lmtest,newdata=data.frame(t=range(P100$t)))
       1        2
33.89501 36.74620

As expected, predict gives two values.  But inside xyplot() predict
gives

300 values:
xyplot(t~s|factor(lonLabels[whichLon100])*factor(latLabels[whichLat
100]),
+   data=P100,pch=".",
+   panel=function(x,y,...){panel.xyplot(x,y,...)
+                           thislm <- lm(x~y)
+                      print(thislm)
+                           newt <- range(P100$t)
+                      print(newt)
+                           news <-
as.vector(predict(thislm,newdata=data.frame(t=newt)))
Why 't' here? The variables involved in 'thislm' are 'x' and 'y'.

Deepayan
+                      print(news)
+                           llines(news,newt,col="red")})
[...]
If you only want a simple linear regression, you might also want to try

 xyplot(t~s|factor(lonLabels[whichLon100])*factor(latLabels[whichLat100]),
+   data=P100,pch=".", type =c("p","r")

 - type "r" will automatically fit and plot the regression line for you.

Hadley
On Tue, 30 Nov 2004, Carlisle Thacker wrote:
Dear R-listers,

It seems that predict() behaves differently within panel.xyplot. 
Am I doing something stupid?
You fitted the model lm(x ~ y) and supplied new values for t, not y.

Using panel.abline would be a bit easier: just call
panel.abline(thislm) from your panel function.
That wouldn't quite work, since the response and predictor are switched 
in the lm fit.

Deepayan
The axes are switched in xyplot.  Oceanographers like to have temperature
as the vertical axis and salinity horizontal.  But temperature is used to
predict salinity.  Confusing, and makes everything complicated.

So panel.abline(thislm) would give the wrong line, as would
type=c("p","r").  And neither would handle a parabolic fit.

The problem was my using t rather than y when getting news.

Thanks for all the help.
On Tuesday 30 November 2004 10:24, Prof Brian Ripley wrote:

On Tue, 30 Nov 2004, Carlisle Thacker wrote:

Dear R-listers,

It seems that predict() behaves differently within panel.xyplot. 
Am I doing something stupid?
You fitted the model lm(x ~ y) and supplied new values for t, not y.

Using panel.abline would be a bit easier: just call
panel.abline(thislm) from your panel function.

That wouldn't quite work, since the response and predictor are switched 
in the lm fit.

Deepayan