scatterplot - set up meaningful threshold for min observations

[cicdguy]

In tm_g_scatterplot, we have 2 validation steps to have at least 10 rows per dataset to be able to perform/display the plot. In a case when selecting MULTIPLE only in the row facetting, the plot is not rendered since only one entry is present for MULTIPLE. Is there a reason why we need at least 10 entries here?

library(random.cdisc.data)
#devtools::load_all("../teal/")
#devtools::load_all("../teal.devel/")
devtools::load_all("../teal.modules.general/")
ADSL <- radsl(cached = TRUE)

app <- init(
  data = cdisc_data(cdisc_dataset("ADSL", ADSL)),
  modules = root_modules(
    tm_g_scatterplot(
      label = "Scatterplot Choices",
      x = data_extract_spec(
        dataname = "ADSL",
        filter = filter_spec(vars = "COUNTRY", choices = value_choices("ADSL", "COUNTRY"), selected = c("CHN", "USA")),
        select = select_spec(
          label = "Select variable:",
          choices = variable_choices(ADSL, c("AGE", "BMRKR1", "BMRKR2")),
          selected = "AGE",
          multiple = FALSE,
          fixed = FALSE
        )
      ),
      y = data_extract_spec(
        dataname = "ADSL",
        filter = filter_spec(vars = "COUNTRY", choices = value_choices("ADSL", "COUNTRY"), selected = c("USA")),
        select = select_spec(
          label = "Select variable:",
          choices = variable_choices(ADSL, c("AGE", "BMRKR1", "BMRKR2")),
          selected = "BMRKR1",
          multiple = FALSE,
          fixed = FALSE
        )
      ),
      color_by = data_extract_spec(
        dataname = "ADSL",
        select = select_spec(
          choices = variable_choices(ADSL),
          selected = "REGION1",
          multiple = TRUE
        )
      ),
      row_facet = data_extract_spec(
        dataname = "ADSL",
        filter = list(
          filter_spec(
            vars = choices_selected(variable_choices("ADSL"), "RACE"),
            multiple = TRUE
          )
        )
      )
    )
  )
)
shinyApp(app$ui, app$server)

Provenance:

Creator: mhallal1

[cicdguy]

I have no idea why 10 points are needed to display visualization. I would expect this only for some statistical method, but it should be based on degrees of freedom or minimal sample size formula and not set a priori to 10.
Let's move it to backlog and remove limitation for all modules instead

Provenance:

Creator: gogonzo

[pawelru]

I would disagree with promotion of this task to validate all modules. Let's do it only for scatterplot

[donyunardi]

Here is the related line:

teal.modules.general/R/tm_g_scatterplot.R

Line 525 in bf8a4a0

teal::validate_has_data(ANL[, c(x_var, y_var)], 10, complete = TRUE, allow_inf = FALSE)

Looking around other modules, the value of min_nrow varied: some 1, 3, or 10.

I want to believe the 10-row restriction was chosen to ensure a meaningful plot, but ultimately, this should be up to the developer. If 10 is too high, we can lower it—even down to 1 if needed.

@kumamiao would like to hear your take on this and decide.

[kumamiao]

For scatter plot, I think having less than 10 data points should not be a reason to not plot the data. Setting to 1 for scatterplot makes sense. Since it's such a quick fix, @donyunardi let's add this to current PI and consider it done, or do you prefer to hold till the backlog-cleaning PI?

[donyunardi]

Sure, I'll raise the PR.