我正在尝试使用rms package中的R来评估线性模型的预测准确性，但由于某种原因而难以生成校准图。
具体来说，我在使用"Error in !fail : invalid argument type"
calibrate()
这是一个简单的可重现示例，用于演示我的问题：
> library(rms)
> library(MASS)
> data(whiteside)
> w.Before <- whiteside[whiteside$Insul=="Before",]
# For comparability with an example in MASS.
> dd <- datadist(w.Before)
> options(datadist="dd")
> mod1 <- ols(Gas ~ Temp,data=w.Before,
x=TRUE,y=TRUE)
> mod1
Coef S.E. t Pr(>|t|)
Intercept 6.8538 0.1184 57.88 <0.0001
Temp -0.3932 0.0196 -20.08 <0.0001
# ols() estimates the coefficients correctly
# Five-fold cross-validation for this model fit:
> validate(mod1,bw=FALSE,method="crossvalidation",B=5)
index.orig training test optimism index.corrected n
R-square 0.9438 0.9431 0.8460 0.0971 0.8467 5
MSE 0.0731 0.0709 0.0929 -0.0220 0.0951 5
g 1.2791 1.2664 1.2222 0.0442 1.2349 5
Intercept 0.0000 0.0000 -0.1661 0.1661 -0.1661 5
Slope 1.0000 1.0000 1.0360 -0.0360 1.0360 5
# try using default argument options:
> calibrate(mod1)
Error in !fail : invalid argument type
# try using some arguments specific to the fitted ols object:
> calibrate(mod1,predy=median(w.Before$Gas),method="crossvalidation",B=5)
Error in !fail : invalid argument type
我做了一些基本的调试（下面） - 也许这可能提供一个线索？
mod1 <- ols(Gas ~ Temp,data=w.Before,
x=TRUE,y=TRUE)
# Switch on debug argument:
> calibrate(mod1, debug = TRUE)
Subscripts of training sample:
[1] 3 9 15 22 21 4 7 22 8 5 11 10 25 4 10 5 18 23 6 1 19 10 9 6 22 25
Subscripts of test sample:
[1] 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26
Error in !fail : invalid argument type
> traceback()
3: predab.resample(fit, method = method, fit = fitit, measure = cal.error,
pr = pr, B = B, bw = bw, rule = rule, type = type, sls = sls,
aics = aics, force = force, estimates = estimates, non.slopes.in.x =
model ==
"ol", smoother = smoother, predy = predy, model = model,
kint = kint, penalty.matrix = penalty.matrix, ...)
2: calibrate.default(mod1, debug = TRUE)
1: calibrate(mod1, debug = TRUE)
> options(error = recover)
> calibrate(mod1)
Error in !fail : invalid argument type
Enter a frame number, or 0 to exit
1: calibrate(mod1)
2: calibrate.default(mod1)
3: predab.resample(fit, method = method, fit = fitit, measure = cal.error, pr = p
Selection: 3
Called from: calibrate.default(mod1)
# Entering object names in predab.resample function from last object in this function until I find an error / missing or suspect values:
Browse[1]> res
Error during wrapup: object 'res' not found
Browse[1]> varin
Error during wrapup: object 'varin' not found
Browse[1]> index.orig
[1] -0.0146323 -0.0095743 -0.0052462 -0.0023136 0.0004922 0.0032980 0.0061038
[8] 0.0081017 0.0111468 0.0147536 0.0183605 0.0219674 0.0255743 0.0291812
[15] 0.0346582 0.0427575 0.0513092 0.0584638 0.0626379 0.0668120 0.0709861
[22] 0.0751603 0.0793344 0.0832308 0.0830547 0.0828785 0.0827024 0.08 25263
[29] 0.0823502 0.0821740 0.0819979 0.0818218 0.0816456 0.0814695 0.0812934
[36] 0.0820341 0.0833151 0.0845299 0.0857447 0.0858007 0.0842924 0.0827842
[43] 0.0801835 0.0750078 0.0698321 0.0646565 0.0591428 0.0513195 0.0434962
[50] 0.0356728
attr(,"keepinfo")
attr(,"keepinfo")$orig.cal
[1] 3.851 3.893 3.935 3.976 4.017 4.057 4.098 4.137 4.178 4.219 4.261 4.302 4.343
[14] 4.385 4.428 4.474 4.520 4.565 4.607 4.649 4.691 4.732 4.774 4.816 4.854 4.891
[27] 4.929 4.966 5.004 5.041 5.079 5.116 5.154 5.191 5.229 5.267 5.306 5.345 5.384
[40] 5.422 5.458 5.494 5.530 5.562 5.595 5.627 5.659 5.689 5.719 5.749
Browse[1]> optimism
Error during wrapup: object 'optimism' not found
希望这提供了一些线索。有人可以帮我解决这个错误，理想情况下，获取这个ols模型的校准图？
提前谢谢。
罗斯。
答案 0 :(得分：0)
我遇到了同样的错误，并且在Github上四处寻找之后，发现这是一个已知问题。
https://github.com/harrelfe/rms/issues/61
好像有一个提交可以在4月份解决该问题，但是尚未将更新的软件包发布到CRAN。
https://github.com/harrelfe/rms/commit/6bcaee45455c0e0c4ec163cd1ac325010f7648fa