[R] newdata for predict.lm() ??

[peter dalgaard]

Don't use $ notation in lm() formulas. Use lm(w ~ h, data=DAT).

-pd

> On 4 Nov 2020, at 10:50 , Boris Steipe <boris.steipe using utoronto.ca> wrote:
> 
> Can't get data from a data frame into predict() without a detour that seems quite unnecessary ...
> 
> Reprex:
> 
> # Data frame with simulated data in columns "h" (independent) and "w" (dependent)
> DAT <- structure(list(h = c(2.174, 2.092, 2.059, 1.952, 2.216, 2.118, 
>                            1.755, 2.060, 2.136, 2.126, 1.792, 1.574,
>                            2.117, 1.741, 2.295, 1.526, 1.666, 1.581,
>                            1.522, 1.995), 
>                      w = c(90.552, 89.518, 84.124, 94.685, 94.710, 82.429,
>                            87.176, 90.318, 76.873, 84.183, 57.890, 62.005,
>                            84.258, 78.317,101.304, 64.982, 71.237, 77.124,
>                            65.010, 81.413)),
>                 row.names = c( "1",  "2",  "3",  "4",  "5",  "6",  "7", 
>                                "8",  "9", "10", "11", "12", "13", "14",
>                               "15", "16", "17", "18", "19", "20"),
>                 class = "data.frame")
> 
> 
> myFit <- lm(DAT$w ~ DAT$h)
> coef(myFit)
> 
> # (Intercept)       DAT$h 
> #   11.76475    35.92002 
> 
> 
> # Create 50 x-values with seq() to plot confidence intervals
> myNew <- data.frame(seq(min(DAT$h), max(DAT$h), length.out = 50))
> 
> pc <- predict(myFit, newdata = myNew, interval = "confidence")
> 
> # Warning message:
> # 'newdata' had 50 rows but variables found have 20 rows 
> 
> # Problem: predict() was not able to take the single column in myNew
> # as the independent variable.
> 
> # Ugly workaround: but with that everything works as expected.
> xx <- DAT$h
> yy <- DAT$w
> myFit <- lm(yy ~ xx)
> coef(myFit)
> 
> myNew <- data.frame(seq(min(DAT$h), max(DAT$h), length.out = 50))
> colnames(myNew) <- "xx"  # This fixes it!
> 
> pc <- predict(myFit, newdata = myNew, interval = "confidence")
> str(pc)
> 
> # So: specifying the column in newdata to have same name as the coefficient
> # name should work, right?
> # Back to the original ...
> 
> myFit <- lm(DAT$w ~ DAT$h)
> colnames(myNew) <- "`DAT$h`"
> # ... same error
> 
> colnames(myNew) <- "h"
> # ... same error again.
> 
> Bottom line: how can I properly specify newdata? The documentation is opaque. It seems the algorithm is trying to EXACTLY match the text of the RHS of the formula, which is unlikely to result in a useful column name, unless I assign to an intermediate variable. There must be a better way ...
> 
> 
> 
> Thanks!
> Boris
> 
> ______________________________________________
> R-help using r-project.org mailing list -- To UNSUBSCRIBE and more, see
> https://stat.ethz.ch/mailman/listinfo/r-help
> PLEASE do read the posting guide http://www.R-project.org/posting-guide.html
> and provide commented, minimal, self-contained, reproducible code.

-- 
Peter Dalgaard, Professor,
Center for Statistics, Copenhagen Business School
Solbjerg Plads 3, 2000 Frederiksberg, Denmark
Phone: (+45)38153501
Office: A 4.23
Email: pd.mes using cbs.dk  Priv: PDalgd using gmail.com