Calculating and Plotting Weekly/Monthly Averages Based on Variable SelectInput

[zmcclean]

Hello Everyone,

I have the following data frame example:

'''

 dates <- seq(as.Date('2018-12-29'), as.Date('2020-01-01'), by = 'days') 
 Variable1 <- sample.int(20, 369, replace = TRUE)
 Variable2 <- sample (20,369, replace = TRUE)

 df <- data.frame(dates, Variable1, Variable2)

'''

My RShiny app has a selectInput (ID="selectvariable") with choices = "Variable1", "Variable2" and a selectInput (ID= "selectduration") with choices = "daily", "weekly", "monthly"

My goal is to create a ggplot with dates on the x axis and the input of "selectvariable" on the y axis. If the input of "selectduration" == "daily" it should plot the selected variable as it appears in the data frame.
If the input of "selectduration" == "weekly" it should plot THE WEEKLY AVERAGE for each week across the data frame for the selected variable.
If the input of "selectduration" == "monthly" it should plot THE MONTHLY AVERAGE for each month across the data frame for the selected variable.

It is a fairly simple app, however I am stuck on how to create these "weekly" and "monthly" averages and plot them. I do not think that creating additional columns of weekly/monthly averages is an appropriate choice because in reality there are many other variables. I want this to be as automated as possible.

I would love to hear what everyone has to say! Thank you!

[nirgrahamuk]

I recommend studying tidy evaluation.
Chapter 12 Tidy evaluation | Mastering Shiny (mastering-shiny.org)

[zmcclean]

Thank you for the suggestion! I have studied this article (it actually helped me answer an unrelated question so that was nice!) however, the question I have in this post remains. Perhaps I was unclear but I am wondering how you can create time-based averages across the data set. I have ran into apply.weekly() but this has been unsuccessful for me so far.

[andresrcs]

@zmcclean

how you can create time-based averages across the data set.

Do you mean something like this?

library(tsibble)
library(dplyr)

df <- data.frame(dates = seq(as.Date('2018-12-29'), as.Date('2020-01-01'), by = 'days'),
                 Variable1 = sample.int(20, 369, replace = TRUE),
                 Variable2 = sample (20,369, replace = TRUE))

df %>% 
    as_tsibble(index = dates) %>% 
    index_by(year_week = ~ yearweek(.)) %>% # weekly aggregates
    summarise(across(where(is.integer), mean, rm.na = TRUE))
#> # A tsibble: 54 x 3 [1W]
#>    year_week Variable1 Variable2
#>       <week>     <dbl>     <dbl>
#>  1  2018 W52     18         8.5 
#>  2  2019 W01     13.6      12.9 
#>  3  2019 W02     11.6       9.86
#>  4  2019 W03      9.71     12.3 
#>  5  2019 W04      9.43     14   
#>  6  2019 W05      8.57     11.9 
#>  7  2019 W06     11         6.14
#>  8  2019 W07     10.6       9.43
#>  9  2019 W08     12.3       8.14
#> 10  2019 W09     11.1       7.71
#> # … with 44 more rows

^{Created on 2021-03-14 by the reprex package (v1.0.0.9002)}

[zmcclean]

@andresrcs

tsibble

Yes, I think this is very close! However, does produce an error for me currently. Would you mind briefly reviewing this aspect of my applicable code? I have included it below.

'''

 output$plot1 <- renderPlot({

  df_filtered <- filter(df, dates >= format(input$daterange[1]) & dates <= format(input$daterange[2])) %>% 
                              filter (client == input$clientInput)

 #as you can see here, there is also a date range filter and a client filter in this data set (I did not include those in my original post because I did not want to overcomplicate it

    if (input$DurationInput == "Daily") {
            df_filtered1 <- df_filtered
 }

 if(input$DurationInput =="Weekly"){
     df_filtered1<- df_filtered %>% 
                                 as_tsibble(index = dates) %>% 
                                 index_by(year_week = ~ yearweek(.)) %>%
                                summarise(across(where(is.integer), mean, rm.na = TRUE))
 }

 #Here I will also include a monthly average (  if(input$DurationInput == "monthly")  ) 
  #and a 'season' average (an average from August to August each year)

    ggplot(df_filtered1, aes(dates, .data[[input$VariableInput]])) +
      geom_point(size=3)+
     geom_line(color = "Red")+
       theme_bw()+
      theme(axis.line = element_line(colour = "black"))
   })

'''

The error I am currently getting is "Warning: Error in FUN: object 'dates' not found" (I am unsure if there are additional errors behind this one.

Thank you very much for your time and knowledge.

[nirgrahamuk]

you can see from andresrcs example that the dates field gets dropped in favour of the year_week variable that gets made by summarising it. You need to account for that.

[zmcclean]

Oh right! this makes complete sense. However, that seems to create a difficulty when attempting to dynamically plot these values. Is there an easy way around this difficulty?

[nirgrahamuk]

'year_week ' is a programmer choice, you could choose a generic name like newdates or dates if you wanted ?

[andresrcs]

There are many approaches you can take, for example, you could also keep the dates column, is all up to you

library(lubridate)
library(dplyr)

df <- data.frame(dates = seq(as.Date('2018-12-29'), as.Date('2020-01-01'), by = 'days'),
                 Variable1 = sample.int(20, 369, replace = TRUE),
                 Variable2 = sample (20,369, replace = TRUE))

df %>% 
    mutate(dates = ceiling_date(dates, unit = "weeks")) %>%
    group_by(dates) %>% 
    summarise(across(where(is.integer), mean, rm.na = TRUE))
#> # A tibble: 54 x 3
#>    dates      Variable1 Variable2
#>    <date>         <dbl>     <dbl>
#>  1 2018-12-30      9        12   
#>  2 2019-01-06      9        10   
#>  3 2019-01-13     12.7       7.71
#>  4 2019-01-20      9.57     10.6 
#>  5 2019-01-27      9.43      8   
#>  6 2019-02-03      5.14      8.71
#>  7 2019-02-10     10.1      13.7 
#>  8 2019-02-17      9.57     11.1 
#>  9 2019-02-24     14.6      11.1 
#> 10 2019-03-03     10.9      14   
#> # … with 44 more rows

^{Created on 2021-03-14 by the reprex package (v1.0.0.9002)}

[zmcclean]

Thank you for both of your help. This solution works almost perfectly! I appreciate all the time you have put into this!