Not sure how to best visualize my data with the string types I have and my Objectives

[Mfon_Amana]

Hi.

I am trying to create a personal project where I find a relationship between whether some Medicines caused any side effects. I have a file where there are 1000 subjects (553 males, 447 females) and indicated whether they took any medicines in buildup to any side effects experienced.

I filtered the dataset to get a shortened table of those who just took Medicine1, or Medicine2 or both. An example is this:

#### Filter Medicine1
Med1 <- survey %>%
  filter(Medicine1 == 'Yes')%>%
  filter(Medicine2 == 'No')%>%
  filter(Medicine3 == 'No')

Since I am looking for a relationship between medicines taken and propensity of side effects, the string type are not numbers, hence I cannot do a scatterplot. I decided to opt for a stacked bar graph (If you think this is the best approach or not, please let me know). Is it possible to incorporate 2 filtered (pipe-operated) dataset together or I would have to create bar graphs for each different filtered datasets?

Here is an example of the chart I made:

ggplot(data=Med1, aes(x = SideEffects, color = Gender))+
  geom_bar()+
  labs(title="Side Effects on Ingesting Medicine 1", caption= "Data Collected from Kaggle")

[AlexisW]

@Mfon_Amana

(If you think this is the best approach or not, please let me know)

The best representation depends strongly on what you want to show (e.g. a difference between gender? A difference between combinations of medicines? The presence of a particular medicine irrespective of the others?). Very roughly, each statistical test you can do could has its own graph that highlights what you are testing.

@Mfon_Amana

Is it possible to incorporate 2 filtered (pipe-operated) dataset together or I would have to create bar graphs for each different filtered datasets?

Both are possible. You can combine pre-filtered datasets with bind_rows() (after ensuring they have the same columns, including a column identifying what filtering was applied). Or in that case, you could also do the whole processing in a single object, for example:

library(tidyverse)

Med <- tibble(SideEffects = sample(c("Yes","No"), 200, replace = TRUE),
               Gender = sample(c("M","F"), 200, replace = TRUE),
               Medicine1 = sample(c("Yes","No"), 200, replace = TRUE),
               Medicine2 = sample(c("Yes","No"), 200, replace = TRUE))

Med |>
  rowwise() %>%
  mutate(medicines = paste0(c_across(starts_with("Medicine")), collapse = ",")) %>%
  ggplot() +
  geom_bar(aes(x = medicines)) +
  facet_wrap(~Gender)

@Mfon_Amana

hence I cannot do a scatterplot

Technically, you can do a scatterplot when x is a string, and on y you are not plotting a string but a count. So, it is possible to make a scatterplot. Whether it's better suited than a barplot, that's a different question. For example here is a scatterplot that highlights the presence of Medicine1:

Med %>%
  mutate(had_medicine1 = (Medicine1 == "Yes"),
         had_medicine2 = (Medicine2 == "Yes")) %>%
  group_by(Gender, had_medicine1, had_medicine2) %>%
  summarize(proportion_with_side_effect = 100 * mean(SideEffects == "Yes")) %>%
  ggplot() +
  geom_point(aes(x = had_medicine1, y = proportion_with_side_effect, color = had_medicine2)) +
  facet_wrap(~Gender)