Error in eval(object$data)[, vars] : object of type 'closure' is not subsettable

General
1

Trying to generate confidence intervals for a logistic function using a bootstrap procedure, and have now spent several hours trying to figure out what is causing this error. Sometimes it works, but then it will fail, even on the same data. I am totally mystified. Reprex below:


library(tidyverse)
library(lubridate)
library(car)

indata <- tribble(
~Date,       ~Cases,  ~Days, 
"2020-03-10",    21,     0, 
"2020-03-11",    23,     1, 
"2020-03-12",    38,     2, 
"2020-03-13",    51,     3, 
"2020-03-14",    56,     4, 
"2020-03-15",    57,     5, 
"2020-03-16",    64,     6, 
"2020-03-17",    83,     7, 
"2020-03-18",   143,     8, 
"2020-03-19",   194,     9, 
"2020-03-20",   304,    10, 
"2020-03-21",   334,    11, 
"2020-03-22",   352,    12, 
"2020-03-23",   410,    13, 
"2020-03-24",   974,    14, 
"2020-03-25",  1396,    15, 
"2020-03-26",  1731,    16, 
"2020-03-27",  2052,    17, 
"2020-03-28",  2371,    18, 
"2020-03-29",  2877,    19, 
"2020-03-30",  3266,    20, 
"2020-03-31",  3997,    21, 
"2020-04-01",  4669,    22, 
"2020-04-02",  5330,    23, 
"2020-04-03",  6110,    24, 
"2020-04-04",  6812,    25, 
"2020-04-05",  7276,    26 
)

indata$Date <- as_date(indata$Date)


#---------------------------------------------------    
#------------------- Fit Logistic function ---------
#---------------------------------------------------    
logistic_model <- function(df, 
                     indep="Cases", # independent variable
                     r=0.24,
                     projection=10){
    
    print(df)
    Asym <- max(df$Cases)*5
    xmid <- max(df$Days)*2
    scal <- 1/r
    
    ## using a selfStart model
    my_model   <- nls(Cases ~ SSlogis(Days, Asym, xmid, scal), 
                   data=df)
    coeffs <- coef(my_model)
    
    dayseq <- df$Days
    dayseq <- c(dayseq,(dayseq[length(dayseq)]+1):
                    (dayseq[length(dayseq)]-1+projection))
    dateseq <- df$Date
    dateseq <- as_date(c(dateseq,(dateseq[length(dateseq)]+1): 
                             (dateseq[length(dateseq)]-1+projection)))
    foo <- tibble(Days=dayseq)
    predict2 <- function(x) {stats::predict(x, foo)}
    print(summary(foo))
    f.boot <- car::Boot(my_model,f=predict2)
    
    Cases <- predict(my_model, data.frame(Days=dayseq))
    confidence <- confint(f.boot)
    preds <- tibble(dayseq, dateseq,
                    Cases,
                    confidence["2.5 %"],
                    confidence["97.5 %"])
    names(preds) <- c("Days", "Date","Cases","SD_upper","SD_lower")
    
    #  return a tibble
    tribble(~Line, ~r, ~K, ~xmid,
            preds, 1/coeffs[["scal"]], coeffs[["Asym"]], coeffs[["xmid"]])
    
} 

        foo <- logistic_model(indata, 
                        indep="Cases", # independent variable
                        r=0.24,
                        projection=10)
2

This is exactly what happens whenever an object is a function, f is asked to return a subset

mean[1]
#> Error in mean[1]: object of type 'closure' is not subsettable

Created on 2020-04-06 by the reprex package (v0.3.0)

Once the head-fake term closure is translated to function and taking operator precedence into account

f(obj) returns a value which, itself may or may not be subsetable

Here it is

a <- mean(seq(1:10))
a[1]
#> [1] 5.5

Created on 2020-04-06 by the reprex package (v0.3.0)

Here, the subset operator [\ ] is bound to a, after mean has done its work. Until that, it's bound to mean(a) and there's nothing that can be subset.

3

That helps, at least I understand what it is objecting to. However, the error is coming from

f.boot <- car::Boot(my_model,f=predict2)

with a traceback of

Error in eval(object$data)[, vars] : object of type 'closure' is not subsettable
9. na.omit(eval(object$data)[, vars])
8. nls(formula = Cases ~ SSlogis(Days, Asym, xmid, scal), data = na.omit(eval(object$data)[, vars]), algorithm = "default", control = structure(list(maxiter = 50, tol = 1e-05, minFactor = 0.0009765625, printEval = FALSE, warnOnly = FALSE), .Names = c("maxiter", "tol", "minFactor", ...
7. eval(call, parent.frame())
6. eval(call, parent.frame())
5. update.default(object, data = na.omit(eval(object$data)[, vars]), start = coef(object))
4. update(object, data = na.omit(eval(object$data)[, vars]), start = coef(object))
3. Boot.nls(my_model, f = predict2)
2. car::Boot(my_model, f = predict2)

  1. logistic_model(indata, indep = "Cases", r = 0.24, projection = 10)

and I just don't see what it doesn't like about what I have fed it.

4

It appears to be an issue with argument handling in the logistic_model function. Unpacking the pieces seems to run as desired.

# as in OP
library(tidyverse)
library(lubridate)
#> 
#> Attaching package: 'lubridate'
#> The following objects are masked from 'package:dplyr':
#> 
#>     intersect, setdiff, union
#> The following objects are masked from 'package:base':
#> 
#>     date, intersect, setdiff, union
library(car)
#> Loading required package: carData
#> 
#> Attaching package: 'car'
#> The following object is masked from 'package:dplyr':
#> 
#>     recode
#> The following object is masked from 'package:purrr':
#> 
#>     some

indata <- tribble(
  ~Date,       ~Cases,  ~Days, 
  "2020-03-10",    21,     0, 
  "2020-03-11",    23,     1, 
  "2020-03-12",    38,     2, 
  "2020-03-13",    51,     3, 
  "2020-03-14",    56,     4, 
  "2020-03-15",    57,     5, 
  "2020-03-16",    64,     6, 
  "2020-03-17",    83,     7, 
  "2020-03-18",   143,     8, 
  "2020-03-19",   194,     9, 
  "2020-03-20",   304,    10, 
  "2020-03-21",   334,    11, 
  "2020-03-22",   352,    12, 
  "2020-03-23",   410,    13, 
  "2020-03-24",   974,    14, 
  "2020-03-25",  1396,    15, 
  "2020-03-26",  1731,    16, 
  "2020-03-27",  2052,    17, 
  "2020-03-28",  2371,    18, 
  "2020-03-29",  2877,    19, 
  "2020-03-30",  3266,    20, 
  "2020-03-31",  3997,    21, 
  "2020-04-01",  4669,    22, 
  "2020-04-02",  5330,    23, 
  "2020-04-03",  6110,    24, 
  "2020-04-04",  6812,    25, 
  "2020-04-05",  7276,    26 
)

indata$Date <- as_date(indata$Date)


# Unpack function to inspect steps
# logistic_model <- function(df., 
#                           indep="Cases", # independent variable
#                           r=0.24,
#                           projection=10){

df. <- indata           # avoid potential namespace confict with df.()
df.
#> # A tibble: 27 x 3
#>    Date       Cases  Days
#>    <date>     <dbl> <dbl>
#>  1 2020-03-10    21     0
#>  2 2020-03-11    23     1
#>  3 2020-03-12    38     2
#>  4 2020-03-13    51     3
#>  5 2020-03-14    56     4
#>  6 2020-03-15    57     5
#>  7 2020-03-16    64     6
#>  8 2020-03-17    83     7
#>  9 2020-03-18   143     8
#> 10 2020-03-19   194     9
#> # … with 17 more rows
r <- 0.24
projection <- 10
indep <- df.$Cases
Asym <- max(indep)
Asym
#> [1] 7276
Asym <- max(indep)*5
Asym
#> [1] 36380
xmid <- max(indep)*2
xmid
#> [1] 14552
scal <- 1/r
scal
#> [1] 4.166667

## using a selfStart model
my_model   <- nls(Cases ~ SSlogis(Days, Asym, xmid, scal), data=df.)
my_model
#> Nonlinear regression model
#>   model: Cases ~ SSlogis(Days, Asym, xmid, scal)
#>    data: df.
#>      Asym      xmid      scal 
#> 10130.468    22.497     3.706 
#>  residual sum-of-squares: 357740
#> 
#> Number of iterations to convergence: 0 
#> Achieved convergence tolerance: 1.495e-06

coeffs <- coef(my_model)
coeffs
#>         Asym         xmid         scal 
#> 10130.467944    22.496894     3.705809

dayseq <- df.$Days
dayseq
#>  [1]  0  1  2  3  4  5  6  7  8  9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24
#> [26] 25 26
dayseq <- c(dayseq,(dayseq[length(dayseq)]+1):
              (dayseq[length(dayseq)]-1+projection))
dayseq
#>  [1]  0  1  2  3  4  5  6  7  8  9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24
#> [26] 25 26 27 28 29 30 31 32 33 34 35
dateseq <- df.$Date
dateseq
#>  [1] "2020-03-10" "2020-03-11" "2020-03-12" "2020-03-13" "2020-03-14"
#>  [6] "2020-03-15" "2020-03-16" "2020-03-17" "2020-03-18" "2020-03-19"
#> [11] "2020-03-20" "2020-03-21" "2020-03-22" "2020-03-23" "2020-03-24"
#> [16] "2020-03-25" "2020-03-26" "2020-03-27" "2020-03-28" "2020-03-29"
#> [21] "2020-03-30" "2020-03-31" "2020-04-01" "2020-04-02" "2020-04-03"
#> [26] "2020-04-04" "2020-04-05"
dateseq <- as_date(c(dateseq,(dateseq[length(dateseq)]+1): 
                       (dateseq[length(dateseq)]-1+projection)))
dateseq
#>  [1] "2020-03-10" "2020-03-11" "2020-03-12" "2020-03-13" "2020-03-14"
#>  [6] "2020-03-15" "2020-03-16" "2020-03-17" "2020-03-18" "2020-03-19"
#> [11] "2020-03-20" "2020-03-21" "2020-03-22" "2020-03-23" "2020-03-24"
#> [16] "2020-03-25" "2020-03-26" "2020-03-27" "2020-03-28" "2020-03-29"
#> [21] "2020-03-30" "2020-03-31" "2020-04-01" "2020-04-02" "2020-04-03"
#> [26] "2020-04-04" "2020-04-05" "2020-04-06" "2020-04-07" "2020-04-08"
#> [31] "2020-04-09" "2020-04-10" "2020-04-11" "2020-04-12" "2020-04-13"
#> [36] "2020-04-14"
foo <- tibble(Days=dayseq)
foo
#> # A tibble: 36 x 1
#>     Days
#>    <dbl>
#>  1     0
#>  2     1
#>  3     2
#>  4     3
#>  5     4
#>  6     5
#>  7     6
#>  8     7
#>  9     8
#> 10     9
#> # … with 26 more rows
predict2 <- function(x) {stats::predict(x, foo)}
print(summary(foo))
#>       Days      
#>  Min.   : 0.00  
#>  1st Qu.: 8.75  
#>  Median :17.50  
#>  Mean   :17.50  
#>  3rd Qu.:26.25  
#>  Max.   :35.00
f.boot <- car::Boot(my_model,f=predict2)
#> Loading required namespace: boot
f.boot
#> 
#> ORDINARY NONPARAMETRIC BOOTSTRAP
#> 
#> 
#> Call:
#> boot::boot(data = data.frame(update(object, model = TRUE)$model), 
#>     statistic = boot.f, R = R, .fn = f, parallel = p_type, ncpus = ncores, 
#>     cl = cl2)
#> 
#> 
#> Bootstrap Statistics :
#>        original       bias    std. error
#> t1*    23.34271   -0.4866764    6.170080
#> t2*    30.55162   -0.6494386    7.634477
#> t3*    39.97806   -0.8572582    9.413758
#> t4*    52.29787   -1.1194215   11.561217
#> t5*    68.38849   -1.4456372   14.132285
#> t6*    89.38583   -1.8450955   17.180618
#> t7*   116.75519   -2.3248455   20.751736
#> t8*   152.37769   -2.8872076   24.873422
#> t9*   198.65316   -3.5259009   29.542128
#> t10*  258.61775   -4.2206262   34.705157
#> t11*  336.07050   -4.9301107   40.239620
#> t12*  435.69562   -5.5842901   45.931675
#> t13*  563.15564   -6.0776290   51.463497
#> t14*  725.11438   -6.2677814   56.419930
#> t15*  929.12766   -5.9866532   60.328484
#> t16* 1183.31907   -5.0732167   62.738629
#> t17* 1495.74780   -3.4361798   63.324491
#> t18* 1873.39421   -1.1456286   61.971226
#> t19* 2320.76197    1.4664003   58.829197
#> t20* 2838.23972    3.7410084   54.471486
#> t21* 3420.56561    4.7221588   50.586974
#> t22* 4055.91561    3.3230391   51.475973
#> t23* 4726.15538   -1.3874649   63.485987
#> t24* 5408.53856   -9.8738393   89.480732
#> t25* 6078.62835  -21.9394712  127.932482
#> t26* 6713.69899  -36.7506952  176.224488
#> t27* 7295.65160  -53.0699325  231.491267
#> t28* 7812.70357  -69.5831710  290.621872
#> t29* 8259.63192  -85.1818215  350.520681
#> t30* 8636.85592  -99.1149886  408.479233
#> t31* 8948.89958 -111.0078632  462.444355
#> t32* 9202.75391 -120.7925148  511.108201
#> t33* 9406.48122 -128.6062318  553.848349
#> t34* 9568.20315 -134.6962370  590.584056
#> t35* 9695.47071 -139.3483623  621.609453
#> t36* 9794.94166 -142.8420482  647.442477
Cases <- predict(my_model, data.frame(Days=dayseq))
Cases
#>  [1]   23.34271   30.55162   39.97806   52.29787   68.38849   89.38583
#>  [7]  116.75519  152.37769  198.65316  258.61775  336.07050  435.69562
#> [13]  563.15564  725.11438  929.12766 1183.31907 1495.74780 1873.39421
#> [19] 2320.76197 2838.23972 3420.56561 4055.91561 4726.15538 5408.53856
#> [25] 6078.62835 6713.69899 7295.65160 7812.70357 8259.63192 8636.85592
#> [31] 8948.89958 9202.75391 9406.48122 9568.20315 9695.47071 9794.94166
#> attr(,"gradient")
#>              Asym        xmid       scal
#>  [1,] 0.002304208   -6.284438   38.15101
#>  [2,] 0.003015816   -8.219390   47.67957
#>  [3,] 0.003946319  -10.745372   59.43284
#>  [4,] 0.005162434  -14.039551   73.86448
#>  [5,] 0.006750773  -18.329821   91.49009
#>  [6,] 0.008823465  -23.907639  112.87939
#>  [7,] 0.011525152  -31.142883  138.63664
#>  [8,] 0.015041525  -40.500121  169.36279
#>  [9,] 0.019609475  -52.554703  205.59075
#> [10,] 0.025528707  -68.005554  247.68245
#> [11,] 0.033174233  -87.679007  295.67508
#> [12,] 0.043008440 -112.514452  349.06464
#> [13,] 0.055590289 -143.517840  406.52168
#> [14,] 0.071577580 -181.664114  465.55152
#> [15,] 0.091716164 -227.726715  522.14509
#> [16,] 0.116807938 -282.016183  570.52203
#> [17,] 0.147648441 -344.028278  603.13834
#> [18,] 0.184926720 -412.043283  611.19138
#> [19,] 0.229087342 -482.783908  585.84463
#> [20,] 0.280168669 -551.311217  520.23111
#> [21,] 0.337651294 -611.366485  411.92558
#> [22,] 0.400368041 -656.282299  265.09332
#> [23,] 0.466528833 -680.355590   91.22562
#> [24,] 0.533888325 -680.278775  -92.35562
#> [25,] 0.600034311 -656.062687 -266.10432
#> [26,] 0.662723482 -611.033443 -412.72542
#> [27,] 0.720169260 -550.904750 -520.77100
#> [28,] 0.771208558 -482.345390 -586.12103
#> [29,] 0.815325803 -411.608108 -611.23583
#> [30,] 0.852562386 -343.622014 -603.00209
#> [31,] 0.883364878 -281.654045 -570.26153
#> [32,] 0.908423378 -227.415172 -521.81197
#> [33,] 0.928533733 -181.403348 -465.18733
#> [34,] 0.944497648 -143.304155 -406.15662
#> [35,] 0.957060499 -112.342194 -348.71853
#> [36,] 0.966879488  -87.541887 -295.35942
confidence <- confint(f.boot)
confidence
#> Bootstrap bca confidence intervals
#> 
#>          2.5 %      97.5 %
#> V1    12.29421    36.78035
#> V2    16.62235    46.77471
#> V3    22.50012    59.62103
#> V4    30.35387    76.08839
#> V5    41.50583    97.16222
#> V6    56.45748   123.92827
#> V7    76.58277   157.73101
#> V8   103.19305   200.37641
#> V9   138.81629   254.28667
#> V10  186.76832   322.77210
#> V11  252.09683   409.49938
#> V12  339.23453   518.00073
#> V13  453.94899   655.36019
#> V14  601.14273   825.64160
#> V15  793.03875  1036.14563
#> V16 1038.62081  1288.26522
#> V17 1346.21539  1606.07691
#> V18 1725.10013  1976.35854
#> V19 2180.92446  2421.26546
#> V20 2716.17593  2935.47340
#> V21 3308.69274  3515.97235
#> V22 3974.49671  4167.46224
#> V23 4650.81530  4848.08154
#> V24 5297.29771  5530.97005
#> V25 5867.97492  6179.78124
#> V26 6421.29890  6796.79802
#> V27 7071.33555  7490.84967
#> V28 7477.23413  8129.73858
#> V29 7754.22767  8669.90743
#> V30 7955.41737  9126.02168
#> V31 8151.33134  9552.85961
#> V32 8284.31206  9913.10172
#> V33 8360.49797 10191.01474
#> V34 8422.28792 10417.73427
#> V35 8501.58178 10634.02729
#> V36 8551.54545 10792.10457
preds <- tibble(dayseq, dateseq,Cases, confidence["2.5 %"], confidence["97.5 %"])
preds
#> # A tibble: 36 x 5
#>    dayseq dateseq    Cases `confidence["2.5 %"]` `confidence["97.5 %"]`
#>     <dbl> <date>     <dbl>                 <dbl>                  <dbl>
#>  1      0 2020-03-10  23.3                    NA                     NA
#>  2      1 2020-03-11  30.6                    NA                     NA
#>  3      2 2020-03-12  40.0                    NA                     NA
#>  4      3 2020-03-13  52.3                    NA                     NA
#>  5      4 2020-03-14  68.4                    NA                     NA
#>  6      5 2020-03-15  89.4                    NA                     NA
#>  7      6 2020-03-16 117.                     NA                     NA
#>  8      7 2020-03-17 152.                     NA                     NA
#>  9      8 2020-03-18 199.                     NA                     NA
#> 10      9 2020-03-19 259.                     NA                     NA
#> # … with 26 more rows
names(preds) <- c("Days", "Date","Cases","SD_upper","SD_lower")
preds
#> # A tibble: 36 x 5
#>     Days Date       Cases SD_upper SD_lower
#>    <dbl> <date>     <dbl>    <dbl>    <dbl>
#>  1     0 2020-03-10  23.3       NA       NA
#>  2     1 2020-03-11  30.6       NA       NA
#>  3     2 2020-03-12  40.0       NA       NA
#>  4     3 2020-03-13  52.3       NA       NA
#>  5     4 2020-03-14  68.4       NA       NA
#>  6     5 2020-03-15  89.4       NA       NA
#>  7     6 2020-03-16 117.        NA       NA
#>  8     7 2020-03-17 152.        NA       NA
#>  9     8 2020-03-18 199.        NA       NA
#> 10     9 2020-03-19 259.        NA       NA
#> # … with 26 more rows

#  return a tibble
tribble(~Line, ~r, ~K, ~xmid,
        preds, 1/coeffs[["scal"]], coeffs[["Asym"]], coeffs[["xmid"]])
#> # A tibble: 1 x 4
#>   Line                  r      K  xmid
#>   <list>            <dbl>  <dbl> <dbl>
#> 1 <tibble [36 × 5]> 0.270 10130.  22.5

#}

Created on 2020-04-06 by the reprex package (v0.3.0)

5

You are correct there. And now I think I'm going crazy. I agree that works, but what could possibly be wrong with the function call? I've tried fiddling it in multiple ways to no avail.

1 Like
6

I have to go all Bill Clinton here: I feel your pain. I've got a similar hang-up about how to reliably pass data frames to functions and get them unpacked. I think I'm finally going to have to get initiated into the mysteries of quasiquotation and the other magic of tidyeval

7

So it is something funky about passing a data frame as a function argument. Totally mysterious to me, so sadly I'll just have to make it a global for now, since that works. I was trying to be good, for my shiny app, and isolate data to each function, but c'est la vie. Truly odd how sometimes it works just fine for other functions I have in the same app. I'll mark this as the "solution", but not very satisfying. Thanks for your help! Sometimes R's eccentricities drive me to distraction.

1 Like
8

This topic was automatically closed 7 days after the last reply. New replies are no longer allowed.