Slider::slide_mean does not behave as I would have expected

[Leo_Carlsson]

Posted this on the R4DS-slack, and was redirected here.

Does anyone have a good grasp of the Slider-package and it's slide_mean function? It does not behave as I thought it would.

library(slider)
library(tidyverse)
tribble(~date, ~value,
        1, 0.557,
        2, 0.674,
        3, 0.586,
        4, 0.851,
        5, 0.151,
        6, 0.154) %>% 
  mutate(roll_avg = slide_mean(value, before = 4, after = 0))
#> # A tibble: 6 x 3
#>    date value roll_avg
#>   <dbl> <dbl>    <dbl>
#> 1     1 0.557    0.557
#> 2     2 0.674    0.616
#> 3     3 0.586    0.606
#> 4     4 0.851    0.667
#> 5     5 0.151    0.564
#> 6     6 0.154    0.483

(.557+.674+.586+.851)/4
#> [1] 0.667
(.674+.586+.851+.151)/4
#> [1] 0.5655

^{Created on 2021-06-10 by the reprex package (v2.0.0)}

I would expect the rolling average for date == 4 to be equal to

(.557+.674+.586+.851)/4

which my calculator gets to 0.667, as does slide_mean. Then I expect it to "jump" forward one step, and take the average of date 2 to 5, ie:

(.674+.586+.851+.151)/4

which my calculator gets to 0.5655. slide_mean however get it to 0.5638. Anyone know what's wrong? Or what I should do to get it to behave as I expected?

I'm on an M1 Macbook Pro, using R 4.1.0 and Rstudio 1.4.1717, if that's something that could be causing the issue

[martin.R]

I think you need to change the 4 to a 3, i.e. 3 before the current row's value:

tribble(~date, ~value,
        1, 0.557,
        2, 0.674,
        3, 0.586,
        4, 0.851,
        5, 0.151,
        6, 0.154) %>% 
  mutate(roll_avg = slide_mean(value, before = 3, after = 0))
# A tibble: 6 x 3
   date value roll_avg
  <dbl> <dbl>    <dbl>
1     1 0.557    0.557
2     2 0.674    0.616
3     3 0.586    0.606
4     4 0.851    0.667
5     5 0.151    0.566
6     6 0.154    0.436

[Leo_Carlsson]

Thank you very much! That was a silly mistake of me