idenfity continuous non-zero values in timeseries

idenfity_continuous_non_zero_series.R

library(tidyverse)

identify_continuous_non_zero_series = function(df, min_sequence_size =5){
  # adds a new column called 'continuous_series' to the df data.frame
  # identifying where 'value' column is >0 for at least min_sequence_size
  
  # filter to only continous chunks > 0, assign each an ID, 
  # and flag each chunk >= min_sequence_size.
  # the cumsum() trick is from https://stackoverflow.com/a/42734207/6615512
  temp_df = df %>%
    filter(value>0) %>%
    arrange(date) %>%
    mutate(sequence_id = cumsum( c(1,diff(date))!=1   )) %>%
    group_by(sequence_id) %>%
    mutate(continuous_series = ifelse(n() >= min_sequence_size, 'Yes','No') ) %>%
    ungroup() %>%
    select(date, continuous_series)
  
  # join to original and input No for all dates where value == 0
  df %>%
    left_join(temp_df, by='date') %>%
    mutate(continuous_series = replace_na(continuous_series, 'No'))
  
}

df <- tibble(
  value = c(.5, .8, .7, .5, .2, .06, 0, 0, 0, 0, 0, .1, .3, .2, 0,0.5, 0.67, 0.32, 0.34,0.34,0.33),
  date = as.Date("2005-07-20") + 0:20
)

identify_continuous_non_zero_series(df, min_sequence_size = 5)