Tukukkk

Let's say I have a dataframe containing 365 observations of a variable, and 365 dates, one for each day of the year.

I want to plot this sequence using ggplot, but I want the plot to begin at an arbitrary date mid-way through the year, and plot all 365 observations, with the dates before the starting point appearing at the end of the sequence.

What can I do, either to the dataframe itself, or to the plot aesthetics, to achieve this?

Updated with reproducible data and more context

Below is example code which should illustrate how I ended up with this problem.

#ten years of data ordered by hydro year



dates <- seq(as.Date("2000-10-01"), as.Date("2010-10-01"), by="days")    

values <- runif(3653)    

df = data.frame(Date=as.Date(dates), Val=values)



> head(df)

        Date       Val

1 2000-10-01 0.9868603

2 2000-10-02 0.6461032

3 2000-10-03 0.7823848

4 2000-10-04 0.9914216

5 2000-10-05 0.8171412

6 2000-10-06 0.3213551



#created new df containing the average of all ten years of measurements for each day of the year



df2 <- df %>% mutate(Day=day(dates), Month =month(dates)) %>%    

group_by(Month, Day) %>%    

summarize(Multiyearmean=mean(Val))



> head(df2)

# A tibble: 6 x 3

# Groups:   Month [1]

  Month   Day Multiyearmean

  <dbl> <int>         <dbl>

1     1     1         0.272

2     1     2         0.577

3     1     3         0.269

4     1     4         0.534

5     1     5         0.607

6     1     6         0.649

My values are still associated with the correct month and day dates, but they are now ordered Jan-Dec, instead Oct-Sep.

Is that interpretation correct?

How can I reorder them for plotting?

How can I achieve the creating the multiyearmean as described here without breaking my date sequence?

EDIT: Original answer created fake data since none originally provided in OP. Now uses suggested df example data. (Thanks for adding, btw! Simplifies answering.)

I'd suggest adjusting your dates so they are a continuous range from one Oct 1 through the next Sept 30. That way you can plot in ggplot using dates, but with the alignment you prefer.

For instance, taking your data, we could adjust it to all go into one Oct-Sep year (ending 2020 so we capture Feb 29's).

df2b <- df %>%

  mutate(date_hydro = lubridate::ymd(paste(

    if_else(month(Date) < 10, 2020, 2019), # 2020 is leap year

    month(Date), day(Date))

  )) %>%

group_by(date_hydro) %>%

summarize(multiyearmean = mean(Val))

Then we can plot the daily averages within the hydrological year.

ggplot(df2b, aes(date_hydro, multiyearmean)) + 

  geom_point() +

  scale_x_date(date_labels = "%b", date_breaks = "1 month", 

               minor_breaks = NULL) +

  theme(axis.text.x = element_text(hjust = 0))

Somewhat similar to the other answer but using your simulation:

set.seed(2018 - 11 - 25) # reproducible data



data.frame(

  dates = seq(as.Date("2000-10-01"), as.Date("2010-10-01"), by = "days"),

  values = runif(3653)

) -> xdf



mutate(

  xdf, 

  day = lubridate::day(dates), 

  month = lubridate::month(dates)

) %>%

  group_by(month, day) %>%

  summarize(multi_year_mean = mean(values)) %>% 

  ungroup() %>% 

  mutate(plot_date = case_when( # use "real" date axis and wrap-around 

    month >= 10 ~ as.Date(sprintf("2019-%02s-%02s", month, day)),

    TRUE ~ as.Date(sprintf("2020-%02s-%02s", month, day)) # account for leap year(s)

  )) %>% 

  ggplot(aes(plot_date, multi_year_mean)) +

  geom_point() +

  scale_x_date(expand=c(0,0.75), date_breaks = "1 month", date_labels = "%b") # adjust aesthetics as necessary