使用 R 可视化多元线性回归

Question

I am trying to fit and visualize a multiple linear regression model using four variables.

fit <- lm(general_data$Bacterial.Contigs ~ general_data$Soil.Temperature + general_data$Time.point + general_data$Year)

Here, Time.point and Year are categorical variables, and others are numerical variables.

I created a 3D plot using the following code.

library(plotly)

plot_ly(data = general_data, z = ~Bacterial.Contigs, x = ~Soil.Temperature, y = ~Time.point, color = ~Year, colors = c('#0C4B8E' ,'#BF382A'),opacity = 0.5) %>%
  add_markers( marker = list(size = 4))

And the plot looks like this:

How can I add the regression line for the "fit" model in this plot. I would really appreciate any help. Thanks

Answer 1

You fitted a model with only additive effects , meaning your categorical values only add or decrease your response variables, the slope will not change for the different categories. It's not easy to visualize that on a 3D plot, I suggest you try ggplot2 .

An example with mtcars , you basically placed the fitted values back into the data frame and call a line for the fitted values:

dat = mtcars
dat$am = factor(dat$am)
dat$vs = factor(dat$vs)

fit <- lm(mpg ~ disp + am + vs,data=dat)
dat$fitted= fitted(fit)

library(ggplo2)
g = ggplot(dat) + geom_point(aes(x=disp,y=mpg)) + 
geom_line(aes(x=disp,y = fitted)) + facet_grid(am~vs)
print(g)

Or if you need a plotly:

library(plotly)
ggplotly(g)