Filtering plots using other variables

Question

Summary

How to create plots that get filtered by the variables not participating in the plot.

Details,

• I have an input data with 3 variables and one prediction,
• I want to show the conditional trend lines in pred vs var_x plots,

Till now,

1. I have made 3 plots - p_x (one for each var_x) and 3 filters - f_x,
2. Each plot is affected by two filters, eg plot p1 gets filters from f2 + f3,
3. I don't want the p1 to get affected by f1 because it's not anymore a trendline,
4. Similar, each plot should get filtered only by the variables not participating in the plot.

Example:

---
title: "Untitled"
format: html
editor: visual
---

```{r, eval = T,echo=F, message=F}
# Libraries
library(plotly)
library(crosstalk)
library(tidyverse)
```


```{r input-prep, eval=T, echo=F, warning=F, message=F, fig.height=4, fig.width=5}

## Data to model
inputData <- mtcars %>% 
  select(var1 = vs, var2 = am, var3 = hp,  mpg) 

## The model 
lm1 <- lm(mpg ~ var1 + var2 + var3,
          data = inputData)

## The grid of different input variable combinations(to be used for predicitons)
graph_data <- expand_grid(
  var1 = seq(min(inputData$var1), max(inputData$var1), length.out = 7),
  var2 = seq(min(inputData$var2), max(inputData$var2), length.out = 7),
  var3 = seq(min(inputData$var3), max(inputData$var3), length.out = 7))

## The new predicted value for the specific input + 95%CI  
pred <- 
    predict(lm1, 
            newdata = graph_data, 
            se = T, 
            interval="confidence", 
            level=0.95)$fit %>% 
    as_tibble() %>% 
      rename(Prediction = fit, `Lower limit` = lwr, `Upper limit` = upr)

## Join the input grid with the predictions - This dataset is to be used for the
## dashboard   
graph_data2 <- bind_cols(graph_data, pred)

```



```{r question, eval=T, echo=F, warning=F, message=F, fig.height=4, fig.width=5}

## Turn the data.frame to SharedData class
mtcars_lowcarb_shared <-  SharedData$new(graph_data2, group = "Choose")

## Preaper the filters - One filter for each variable 
f1 <- filter_select(id = "lowcarb_selector",
                label = "Select var1",
                sharedData = mtcars_lowcarb_shared,
                group = ~as.character(var1 %>% round(2)),
                multiple = F
                )
f2 <- filter_select(id = "lowcarb_selector",
                label = "Select var2",
                sharedData = mtcars_lowcarb_shared,
                group = ~as.character(var2 %>% round(2)),
                multiple = F
                )

f3 <- filter_select(id = "lowcarb_selector",
                label = "Select var3",
                sharedData = mtcars_lowcarb_shared,
                group = ~as.character(var3 %>% round(2)),
                multiple = F
                )

# Make the line plot for var1  
p1 <- mtcars_lowcarb_shared %>%
  ggplot(mapping = aes(x = var1, y = Prediction, group =1, fill =1)) +
  geom_line()

p1 <- ggplotly(p1)

## The var1 should not be filtered (by cb1)
bscols(list(f2, f3), list(p1), widths = c(7, 7))


## 2nd variable
# Prepare data and create shared data objects
mtcars_lowcarb_shared2 <-  SharedData$new(graph_data2, group = "Choose")

## Line plot for var2  
p2 <- mtcars_lowcarb_shared2 %>%
  ggplot(mapping = aes(x = var2, y = Prediction, group =1, fill =1) ) +
  #geom_point(data = ) +
  geom_line()

p2 <- ggplotly(p2)

## The var2 should not be filtered (by cb2) - But with this code it does :(
bscols(list(f1, f3), list(p2), widths = c(7))


# Prepare data and create shared data objects
mtcars_lowcarb_shared3 <-  SharedData$new(graph_data2, group = "Choose")

## Line plot for var3  
p3 <- mtcars_lowcarb_shared3 %>%
  ggplot(mapping = aes(x = var3, y = Prediction, group =1, fill =1) ) +
  #geom_point(data = ) +
  geom_line()

p3 <- ggplotly(p3)

## The var2 should not be filtered (by cb3) - But with this code it does :(
bscols(list(f1, f2), list(p3), widths = c(7))

```

The outcome

The plot of var1 get filtered nicely from filters f2 + f3 (which are filtering var2 + var3)

The plot for var2

• Has got filtered by f1 (good) but
• Is affected by the f2 (problem!).
  

The same issue applied for var3 which is affected by f3. Instead it should be affected by f1 + f2