How to graph with geom_ribbon

Question

I have three tables:

Upper  Bound  
    Q      C   
    1     30  
    2     50  
    3     40

Lower  Bound      
    Q      C       
    1     10   
    2     15     
    3     20 

Bad Data:

Q      C      Name  
1      50     Sample 1  
2      40     Sample 1  
3      30     Sample 1  
1      0      Sample 2  
2      60     Sample 2  
3      5      Sample 2

I want a graph that graphs the lower and upper bounds in gray and fills everything in between and graph the bad samples on top with different colors and a legend:

plot <- ggplot(Bad_Data, aes(x = Bad_Data$Q, y = Bad_Data$C, group = 1))
plot + geom_line(aes(color = N)) + geom_ribbon(aes(ymin = Lower_Bound$C, ymax = Upper_Bound$C))  

I tried that but it gave me this error:

Error: Aesthetics must be either length 1 or the same as the data (624): ymin, ymax, x, y, group

Anyone who can help me?

Answer 1

I use geom_ribbon to shade forecast confidence intervals with forecasts. You can combine several data frames to make it all work. You need to pass the data frame, x values (dates) and y values (two lines to shade between).

The following function is used for plotting a series, the model backtest, predictions and confidence intervals.

plot_predictions <- function(start_date) {
    # function to use Plotly to plot the predictions,
    # confidence interval and actuals
    # inputs:
    # plot_df, starting date of the series, r-pred (prediction data frame)

# output: interactive plot of the actual and the model backtest

# set the end date parameter first
pred_end <- as.Date(tail(r_pred$week_ending, 1))


p <- ggplot() + 
# add the backtest series from the backtest data frame
geom_line(data=r_back, mapping = aes(x=week_ending, y=Backtest_Model,
                                   color = 'Backtest Model'), linetype='solid') + 

# add the actual series from the training dataframe
# use aes_string to pass the series variable
geom_line(data=r_train, mapping = aes_string(x='week_ending', y=series,
                                   color = 'series'), linetype='solid') +

# r_pred holds he predictions and confidence levels
geom_line(data=r_pred, mapping = aes(x=week_ending, y=Predictions,
                                   color = 'Predictions'), linetype='solid') + 


# upper forecast limit
geom_line(data=r_pred, mapping = aes(x=week_ending, y=upper_conf_limit,
                                   color = 'upper_limit'), linetype='solid') + 

# lower forecast limit
geom_line(data=r_pred, mapping = aes(x=week_ending, y=lower_conf_limit,
                                   color = 'lower_limit'), linetype='solid') + 


# format the plot 
scale_linetype_manual() + 

# prediction_pal is a five color palette
scale_color_manual(values = pred_pal, name = "Series") +

# fill between lines with geom_ribbon using a blue toned down by alpha
geom_ribbon(data=r_pred, aes(x = week_ending, 
                             ymin=lower_conf_limit,
                             ymax=upper_conf_limit),
            fill="blue", alpha=0.2) + 



# format the y axis with commas
scale_y_continuous(label = comma)

# extend the date series from the beginning to end of predictions
scale_x_date(breaks = pretty_breaks(20),
             limits = c(as.Date(start_date), pred_end)) +  


# add the custom theme    
theme_bryan() + 

# customize the plot
# rotate the text of the x axis
theme(axis.text.x = element_text(angle= 45, hjust = 1)) + 
labs(fill = 'Series') + 
guides(color = guide_legend(reverse = FALSE)) +


# add the holiday lines as vertical red dotted lines
# holidays in the training set
geom_vline(xintercept = as.numeric(holiday$week_ending),
           linetype='dotted', colour = 'red', alpha =0.5) + 

# holidays in the forecasting set
    geom_vline(xintercept = as.numeric(r_exog_lines$week_ending),
               linetype='dotted', colour = 'red', alpha =0.5)

# output the plot in Plotly format
subplot(with_options(list(digits = 0),
                     ggplotly(p))) %>% 
    layout(legend = list(orientation = 'v', y = .1, x = 0))
}

Using a series called riders series <- 'riders' yields the following.

Answer 2

Here is a start, you can tweak the colors and other parameters to get the dynamics exactly as you like. I have added a few aesthetics with description of what each does:

#Prepare
Bad_Data$Lower_Bound <- Lower_Bound$C
Bad_Data$Upper_Bound <- Upper_Bound$C

#Plot
library(ggplot2)
p <- ggplot(Bad_Data, aes(x = Q, y = C, color=Name, group=Name))
p <- p + geom_line()
p + geom_ribbon(aes(ymin=Lower_Bound, ymax=Upper_Bound), 
                alpha=0.1,       #transparency
                linetype=1,      #solid, dashed or other line types
                colour="grey70", #border line color
                size=1,          #border line size
                fill="green")    #fill color

Data

Here is the data:

Upper_Bound <- read.table(text="Q      C   
1     30  
                          2     50  
                          3     40", header=T)

Lower_Bound <- read.table(text="Q      C       
                          1     10   
                          2     15     
                          3     20", header=T) 

Bad_Data <- read.table(text="Q      C      Name  
                       1      50     Sample1  
                       2      40     Sample1  
                       3      30     Sample1  
                       1      0      Sample2  
                       2      60     Sample2  
                       3      5      Sample2", header=T)

Edit

Safer preparation:

Bad_Data$Lower_Bound <- Lower_Bound$C[match(Bad_Data$Q, Lower_Bound$Q)]
Bad_Data$Upper_Bound <- Upper_Bound$C[match(Bad_Data$Q, Upper_Bound$Q)]