通过示例在 R 中使用时间序列进行预测的问题

Question

Im working through this example . However, when I begin investigating the tk_ts output I don't think it is taking the start/end data Im entering correctly, but am unsure as to what the proper input is if I want it to start at 12-31-2019 and end at 7-17-2020:

daily_cases2 <- as_tibble(countrydatescases) %>%
    mutate(Date = as_date(date)) %>%
    group_by(country, Date) %>%
    summarise(total_cases = sum(total_cases))

  daily_cases2$total_cases <- as.double(daily_cases2$total_cases)
  
  # Nest
  daily_cases2_nest <- daily_cases2 %>%
    group_by(country) %>%
    tidyr::nest()
  
# TS
  daily_cases2_ts <- daily_cases2_nest %>%
    mutate(data.ts = purrr::map(.x  = data, 
                         .f       = tk_ts, 
                         select   = -Date,
                         start    = 2019-12-31,
                         freq     = 1))

Here is what I get when I examine it closely:

When I go through the example steps with these parameters the issue is also then seen in the subsequent graph:

I've tried varying the frequency and start parameters and its just not making sense. Any suggestions?

Answer 1

You've given the start and end dates, but you haven't said what frequency you want. Given that you want the series to start at the end of 2019 and end in the middle of July, 2020, I'm guessing you want a daily time series. In that case, the code should be:

daily_cases2_ts <- daily_cases2_nest %>%
    mutate(data.ts = purrr::map(.x  = data, 
                         .f       = tk_ts, 
                         select   = -Date,
                         start    = c(2019, 365),  # day 365 of year 2019
                         freq     = 365))          # daily series