Custom dplyr function and multiple ggplot of multiple argument

Question

I need some help for a technical manipulation on R plz.

My problem : I have some observation data of bird by presence/absence in differents habitat types. I want know the sucess ratio of observation in these differents habitats according to their surface range :

data_observation <- data.frame(
  habitat_bush = c(
    0, 0, 0, 0, 10,
    10, 30, 30, 30, 45,
    65, 65, 65, 80, 80,
    80, 90, 95, 100
  ),
  obs = c(
    "yes", "no", "no", "no", "yes",
    "no", "no", "yes", "no", "yes",
    "yes", "no", "yes", "no", "yes",
    "yes", "yes", "yes", "yes"
  )
)

Here you have just data for 'habitat_bush" but in have 10 more time habitats.

Help by a colleague, we have made this function to make a ggplot of the ratio success of observation under differents area size of 'habitat_bush" :

library(dplyr)
library(ggplot2)
library(scales)


plot_forest_test <- function(data = NULL, habitat_type = NULL, colour = NULL) {
  x <- enquo(habitat_type)
  fill <- enquo(colour)

  ggdata <- data %>%
    select(x = !!x, fill = !!fill) %>%
    mutate(
      group = case_when(
        x == 0 ~ "[0]",
        x > 0.0001 & x < 10.0001 ~ "]0-10]",
        x > 10.0001 & x < 25.0001 ~ "]10-25]",
        x > 25.0001 & x < 50.0001 ~ "]25-50]",
        x > 50.0001 & x < 75.0001 ~ "]50-75]",
        x > 75.0001 ~ "]75- 100]"
      )
    ) %>%
    select(-x) %>%
    group_by(group, fill) %>%
    count() %>%
    group_by(group) %>%
    group_modify(~ mutate(.data = .x, freq = n / sum(n)))

  ggplot(data = ggdata, mapping = aes(x = group, y = freq, fill = fill)) +
    geom_bar(stat = "identity") +
    scale_fill_brewer(palette = "Greens") +
    scale_y_continuous(labels = scales::percent) +
    theme_minimal() +
    labs(x = expr(!!x), fill = expr(!!fill))
}

plot_forest_test(data = data_observation, habitat_type = habitat_bush, colour = obs)

It's work very well. But the observation can depend of effort put by technicien to looking for the presence of bird. So, I have data like that :

data_observation_2 <- data.frame(
  superficie_essence = c(
    0, 0, 0, 0, 10,
    10, 30, 30, 30, 45,
    65, 65, 65, 80, 80,
    80, 90, 95, 100
  ),
  obs = c(
    "yes", "no", "no", "no", "yes",
    "no", "no", "yes", "no", "yes",
    "yes", "no", "yes", "no", "yes",
    "yes", "yes", "yes", "yes"
  ),
  effort = c(low, low, mid-low, mid-low, low, mid-low, mid-low,
            mid-high, mid-high, high, mid-low, mid-low, mid-high, mid-low, mid-high, high, high, mid-high, high)
)

My R skills stop here. I want have the same previously graph but subdivided by effort_type for each modalities of habitats types, in the same graphical (like multipanel graphical). In other word I want 5 sub-graph of previous graph with 1 barplot by efforts modalities. But I have lot of data, so I would like put this processu into a function like :

plot_forest_test_2(data = data_observation, habitat_type = habitat_bush, effort = Q_effort, colour = obs)

Can you help me please ? Thanks for your help !

cdlt

yake84 · Accepted Answer · 2019-06-03 02:29:14Z

1

Quosures are not my forte, especially when they might be missing but give this a shot. I created a new column for the facetted item and then add facet_wrap(). You could also use facet_grid(). Hope it helps.

plot_forest_test <- function(data = NULL, habitat_type = NULL, colour = NULL, facet = NULL) {
  x <- enquo(habitat_type)
  fill <- enquo(colour)

  # this is new ####################
  facet <- enquo(facet)
  has_facet <- quo_name(facet) != "NULL"

  df <- 
    data %>% 
    mutate(
      x = !!x, 
      fill = !!fill,
      facet = ""
    )

  if (has_facet) {
    df <- 
      df %>% 
      mutate(facet = !!facet)
  }
  ##################################

  ggdata <- 
    df %>%
    mutate(
      group = case_when(
        x == 0 ~ "[0]",
        x > 0.0001 & x < 10.0001 ~ "]0-10]",
        x > 10.0001 & x < 25.0001 ~ "]10-25]",
        x > 25.0001 & x < 50.0001 ~ "]25-50]",
        x > 50.0001 & x < 75.0001 ~ "]50-75]",
        x > 75.0001 ~ "]75- 100]"
      )
    ) %>%
    select(-x) %>%
    # adding facet here
    group_by(group, fill, facet) %>% 
    count() %>%
    group_by(group, facet) %>%
    arrange(desc(fill)) %>% 
    mutate(
      freq = n/sum(n),
      # these steps set up the label placement
      running_freq = cumsum(freq),
      prev_freq = lag(running_freq, default = 0),
      label_y = (prev_freq + running_freq)/2 ,
      label_n = paste0("n = ", sum(n))
    ) %>% 
    ungroup()

  # create plot w/o facet
  p <-
    ggplot(data = ggdata, mapping = aes(x = group, y = freq, fill = fill)) +
    geom_bar(stat = "identity") +
    geom_hline(yintercept = 0) +
    geom_text(aes(y = -0.05, label = label_n), size = 3.5) +
    #geom_text(aes(y = label_y, label = n)) +
    scale_fill_brewer(palette = "Greens") +
    scale_y_continuous(labels = scales::percent) +
    theme(
      panel.background = element_rect(fill = "white"),
      panel.border = element_rect(color = "grey90", fill = NA)
    ) +
    labs(x = expr(!!x), fill = expr(!!fill))

  # add in if facet was mentioned
  if (has_facet) {
    p <-
      p +
      facet_grid(~facet)
  }

  # return final plot
  p
}

I am including an edit to data_observation_2 as the strings were not in quotes and some of the values has spaces around the hyphens when others did not. I made them all consistent w/o spaces

data_observation_2 <- data.frame(
  superficie_essence = c(
    0, 0, 0, 0, 10,
    10, 30, 30, 30, 45,
    65, 65, 65, 80, 80,
    80, 90, 95, 100
  ),
  obs = c(
    "yes", "no", "no", "no", "yes",
    "no", "no", "yes", "no", "yes",
    "yes", "no", "yes", "no", "yes",
    "yes", "yes", "yes", "yes"
  ),
  effort = c(
    "low", "low", "mid-low", "mid-low", "low", "mid-low", "mid-low",
    "mid-high", "mid-high", "high", "mid-low", "mid-low", 
    "mid-high", "mid-low", "mid-high", "high", "high", "mid-high", "high"
  )
  )
)

And the final outcome. I used fct_relevel() to put them in order of effort.

plot_forest_test(
  data = data_observation, 
  habitat_type = habitat_bush, 
  colour = obs
)

data_observation_2 %>% 
  mutate(effort = fct_relevel(effort, "low", "mid-low", "mid-high", "high")) %>% 
  plot_forest_test(
    habitat_type = superficie_essence, 
    colour = obs, 
    facet = effort
  )

edited Jun 3, 2019 at 2:29

answered Jun 1, 2019 at 3:20

yake84

3,3163 gold badges24 silver badges40 bronze badges

Sign up to request clarification or add additional context in comments.

5 Comments

thomas leon Over a year ago

It's perfect thx ! Can I ask another question, please ? Is it possible to display on the graph the number "n" with which the frequencies have been calculated ?

yake84 Over a year ago

Sure. Are the barcharts supposed to fill 0-100 for each category or should they look like the chart I submitted above (total bar heights all different)?

yake84 Over a year ago

I made updates to add data labels and put them in order of effort. If the solution works for you, please mark it as answered.

thomas leon Over a year ago

Thank you for your help. I'm not sure I understood what you're telling me. I calculated frequencies on an "n" observation. However when I have 60% failure / 40% success on a surface modality and in a modality of effort, however, it misses the information on the number of observations of the calculation of the frequency (if it is a frequency calculated on 2 observations or 100 or 1000). So I wondered if we could show this information on the graph with for example, on each base of the barplots "n = 145" or "n = 175" etc. Thanks again !

yake84 Over a year ago

I updated the code with my go-to method, I imagine there are a few ways people might do this.

Collectives™ on Stack Overflow

Custom dplyr function and multiple ggplot of multiple argument

1 Answer 1

5 Comments

Your Answer

Hot Network Questions

Collectives™ on Stack Overflow

1 Answer 1

5 Comments

Your Answer

Sign up or log in

Post as a guest

Related