rollify-centered.R

library(tibbletime)
library(dplyr)
library(zoo)

data(FB)

window <- 6
rolling_mean <- rollify(mean, window = window)

FB_roll <- FB %>%
  mutate(
    # Using rollmean, with na.pad and align = "center"
    rollmeaned = rollmean(adjusted, k = window, na.pad = TRUE, align = "center"),
    
    # Rollify then Lead it
    rollified  = rolling_mean(adjusted) %>% lead(n = round(window/2))
  ) 

FB_roll %>% select(date, adjusted, rollmeaned, rollified)
#> # A tibble: 1,008 x 4
#>    date       adjusted rollmeaned rollified
#>    <date>        <dbl>      <dbl>     <dbl>
#>  1 2013-01-02     28.0       NA        NA  
#>  2 2013-01-03     27.8       NA        NA  
#>  3 2013-01-04     28.8       28.9      28.9
#>  4 2013-01-07     29.4       29.5      29.5
#>  5 2013-01-08     29.1       30.1      30.1
#>  6 2013-01-09     30.6       30.5      30.5
#>  7 2013-01-10     31.3       30.6      30.6
#>  8 2013-01-11     31.7       30.8      30.8
#>  9 2013-01-14     31.0       30.7      30.7
#> 10 2013-01-15     30.1       30.4      30.4
#> # ... with 998 more rows

FB_roll %>%
  summarise(`Equal?` = all.equal(rollmeaned, rollified))
#> # A tibble: 1 x 1
#>   `Equal?`
#>   <lgl>   
#> 1 T

#' Created on 2018-01-31 by the [reprex package](http://reprex.tidyverse.org) (v0.1.1.9000).

Note that something like window = 7 doesn't give the same results. zoo shifts up one more row than we do there. Not sure why as I feel like with a window of 7 the center point is clearly to place it at row 4 (3 on either side).