Plotting multiple X variables for one continuous Y variable in a single plot

Question

I am trying to draw a line plot having two x variables in the x-axis with one continuous y variable in the y-axis. The count of x1 and x2 are different. The df looks like the following-

df <- structure(list(val = c(3817,2428,6160,6729,7151,7451,6272,7146,7063,6344,5465,6169,7315,6888,7167,6759,4903,6461,7010,7018,6920,3644,6541,31862,31186,28090,28488,29349,28284,25815,23529,20097,19945,22118), type = c("1wt", "1wt", "3wt", "3wt", "3wt", "5wt", "5wt", "7wt", "7wt", "7wt","10wt","10wt","10wt","15wt","15wt","20wt","20wt","25wt","25wt","25wt","30wt","30wt","30wt","20m","20m","15m","15m","15m","10m","10m","5m", "5m", "5m", "5m"), group = c("A", "A", "A", "A", "A", "A", "A", "A", "A", "A", "A", "A", "A", "A", "A", "A", "A", "A", "A", "A", "A", "A", "A", "B", "B", "B", "B", "B", "B", "B", "B", "B", "B", "B")), row.names = c(NA, 34L), class = "data.frame")

where the x variables are-

x1 <- factor(df$type, levels = c("1wt", "3wt", "5wt", "7wt", "10wt", "15wt", "20wt", "25wt", "30wt")) and

x2 <- factor(df$type, levels = c("20m", "15m","10m","5m"))

I want to have separate lines for the x1 and x2 with different colors and legends as per the df$group at the x-axis with df$val at the y -axis. could you please help me doing this? Thanks in advance.

Answer 1

EDIT: added below

Here's an approach that assumes the intent is to map the span of possible type values from group A against the span of possible values from group B.

Labeling could be added manually, but I don't think there's any simple way to use two categorical x axes together in one plot.

df2 <- df %>%
  mutate(x = case_when(type == "1wt" ~ 0,
                       type == "3wt" ~ 1,
                       type == "5wt" ~ 2,
                       type == "7wt" ~ 3,
                       type == "10wt" ~ 4,
                       type == "15wt" ~ 5,
                       type == "20wt" ~ 6,
                       type == "25wt" ~ 7,
                       type == "30wt" ~ 8,

                       type == "20m"  ~ 0/3 * 8,
                       type == "15m"  ~ 1/3 * 8,
                       type == "10m"  ~ 2/3 * 8,
                       type == "5m"   ~ 3/3 * 8))


ggplot(df2, aes(x, val, color = group, group = group)) + 
  geom_point() +
  geom_smooth(method = lm)

2nd approach

It sounds like the OP would like to use the type values numerically in some fashion. If they aren't intrinsically linked to each other in the way that's described, I suspect it will be misleading to plot them as if they are. (See here for a discussion of why this is trouble.)

That said, here's how you could do it. First, here's an approach that just uses the numeric portion of type as is. Note that "m", associated with group B, is on the bottom and "wt" is on the top, associated with group A, as in the example added in the OP comment below. I've added colors to the axes to clarify this. It's a little counterintuitive visually, since the points related to the top axis are on the bottom, and vice versa.

df2 <- df %>%
  # First, let's take the number used in "type" without adjustment
  mutate(x_unadj = parse_number(type))

ggplot(df2, aes(x_unadj, val, color = group, group = group)) + 
  geom_point() +
  geom_smooth(method = lm) + # Feel free to use other smoothing method, but
                             # not obvious to me what would be improvement.
  scale_x_continuous("m", sec.axis = sec_axis(~., name = "wt")) +
  theme(axis.text.x.bottom  = element_text(color = "#00BFC4"),
        axis.title.x.bottom = element_text(color = "#00BFC4"),
        axis.text.x.top     = element_text(color = "#F8766D"),
        axis.title.x.top    = element_text(color = "#F8766D"))

If this is not satisfactory, we might reverse the order of both axes using

scale_x_reverse("m", sec.axis = sec_axis(~., name = "wt")) +

Using ggplot 3.1.0 (from Oct 2018), I could not get the secondary x axis to shift in the opposite direction as the primary axis. This example from 2017 doesn't seem to work with this version any more. As of Dec 2018, there is a proposed fix being reviewed that is meant to address this.