Add multiple parallel planes to 3D plot with plotly package (datacamp exercise) - matrix

Question

I am trying to visualize two parallel planes in a 3D plot with the plotly package (for those that have access to Datacamp, it is this exercise: https://campus.datacamp.com/courses/multiple-and-logistic-regression/multiple-regression?ex=9)
At some point in the construction of this graph, you need to create a matrix to fit the planes that match the categorical variable (plane_new = condition is new, plane_used = condition is used). The problem is, that the two datasets that I used as input (corresponding to the two levels of condition), have different number of observations. I can't seem to figure out how I can get these matrices to be comparable, so that the planes are correctly fit as geometrical objects in the figure.
I hope an R-wizard can help me out ;). Here's my code:
# libraries
library(openintro) #exemplary datasets
library(modelr) #multivariate methods
library(broom) #tidy
library(ggplot2) #visualizing data
library(plotly) #visualizing models in 3D
## Fit the model
lm_ext <- lm(totalPr ~ duration + startPr + cond, # Interpretation: With every 1 unit increase of auction duration (unit = day), the price of the game decreases with .51 units in the response variable (total price), when keeping startPr constant. The eventual value of the predicted value also depends on condition (categorical), for which the y-intercept is different
data = marioKart)
## Visualize the model (including predictions)
marioKart_ss_new <- subset(marioKart, cond == "new") # To visualize planes in a 3D graph in plotly, the dataframe needs to be split in the number of levels of the categorical variable
marioKart_ss_used <- subset(marioKart, cond == "used")
duration_new <- as.vector(marioKart_ss_new$duration) # These vectors represent the linear model for condition = new
startPr_new <- as.vector(marioKart_ss_new$startPr)
duration_used <- as.vector(marioKart_ss_used$duration) # These vectors represent the linear model for condition = used
startPr_used <- as.vector(marioKart_ss_used$startPr)
lm_new <- lm(totalPr ~ duration + startPr, # Create two linear models
data = marioKart_ss_new)
lm_used <- lm(totalPr ~ duration + startPr,
data = marioKart_ss_used)
grid_new <- marioKart_ss_new %>% # Make two grids with all combinations of the levels of the two numerical explanatory variables
data_grid(duration =
seq_range(duration, by = 1),
startPr =
seq_range(startPr, by = 1))
grid_used <- marioKart_ss_used %>%
data_grid(duration =
seq_range(duration, by = 1),
startPr =
seq_range(startPr, by = 1))
lm_new <- lm(totalPr ~ duration + startPr, # Make two seperate models based on the two levels of the categorical explanatory variable
data = marioKart_ss_new)
lm_used <- lm(totalPr ~ duration + startPr,
data = marioKart_ss_used)
pred_new <- augment(lm_new, newdata = grid_new) # Predictions
pred_used <- augment(lm_used, newdata = grid_used)
plane_new <- matrix(pred_new$.fitted, # Matrix of preditions as input for planes
nrow = 70,
ncol = 70)
plane_used <- matrix(pred_used$.fitted,
nrow = 55,
ncol = 55)
plot <- plot_ly(data = marioKart, # 3D plot of datapoints
z = ~totalPr,
x = ~duration,
y = ~startPr,
opacity = 0.6) %>%
add_markers(color = ~cond)
plot %>% # Add planes
add_surface(x = ~duration_new, ### NOT WORKING, WAIT FOR DATACAMP
y = ~startPr_new,
z = ~plane_new,
showscale = FALSE) %>%
add_surface(x = ~duration_used,
y = ~duration_used,
z = ~plane_used,
showscale = FALSE)

Answer 1

No code wizard here but asked for the same thing:
library(tidyverse)
library(modelr)
grid <- mario_kart %>%
modelr::data_grid(
duration = seq_range(duration, n = 70),
startPr = seq_range(startPr, n = 70),
cond
)
library(broom)
tidy_planes <- mod %>%
augment(newdata = grid)
x <- unique(grid$duration)
y <- unique(grid$startPr)
plane0 <- tidy_planes %>%
filter(cond == "new") %>%
pull(.fitted) %>%
matrix(nrow = length(x), byrow = TRUE)
plane1 <- tidy_planes %>%
filter(cond == "used") %>%
pull(.fitted) %>%
matrix(nrow = length(x), byrow = TRUE)