Stacked bar plot with points, but with different aestetics length - ggplot2

Question

I have a dataframe which I used the melt function to get to this (length = 118): 

 record_id          value Values
1           8     int_to_out     20
2          14     int_to_out     32
3           5     int_to_out     22
4           6     int_to_out     19
5          31     int_to_out     15
6          48     int_to_out     20
7         100     int_to_out     30
...       ...        ...        ...
113        87 symptom_to_int      7
114        72 symptom_to_int      4
115        99 symptom_to_int      3
116       102 symptom_to_int     36
117       103 symptom_to_int     13
118       111 symptom_to_int      6

I made a stacked barplot with this:

The plot has 59 y elements, and I need to add points to them based in the original (non-melted) data. So I wrote this:

ggplot(t, aes(y=as.factor(record_id), x=Values, fill=value)) + 
    geom_bar(position=position_stack(reverse= TRUE), stat="identity") +
    geom_point(data = new_df, aes(x=sorolog, y = record_id), 
                colour = "#a81802", size = 4, shape = 1)

The x = sorolog has 59 values for the 59 IDs foun in record_id.

But when I run it I get this:

    Error: Aesthetics must be either length 1 or the same as the data (59): fill
Run `rlang::last_error()` to see where the error occurred.

Which I believe that is a conflict with the melted data, since it's length is the double of the original dataframe.

The question is: How can I add the points with this difference of aestetics length?

Another problem: How can I add a second legend to the plot?

I used this code:

ggplot() + 
    geom_bar(data=t, aes(y=as.factor(record_id), x=Values, fill=value), 
        position=position_stack(reverse= FALSE), stat="identity", width = 0.5) +
        scale_fill_manual(values = c("brown1","chocolate1"),name = "", 
            labels = c("Hospitalization to Discharge", "Symptom to Hospitalization")) +
    geom_point(data = new_df, aes(x=sorolog, y = as.factor(record_id)), 
                colour = "darkcyan", size = 5, shape = 1)+
    geom_point(data = new_df, aes(x=final, y = as.factor(record_id)), 
                colour = "darkred", size = 4, shape = 16)+

        theme_minimal()+
    labs(title="Patient timeline - from symptoms to hospitalization and discharge",
        x ="Days", y = "Patient ID")+
    theme(text = element_text(family = "Garamond", color = "grey20"))

and got this:

but I can't add a legend for the geom_point elements, how can I do it?

EDIT

With the edit from Dave Armstrong I got this:

In the bar plot you set `y=as.factor(record_id)`, whereas your `new_df` has `y` mapped just to `record_id`. If you did not set `new_df$record_id` to be factored beforehand, it is likely `ggplot2` is trying to use this as a continuous aesthetic rather than discrete (`as.factor`). Is this what's going on here? — chemdork123, Sep 01 '20 at 20:23
Is there something else named sorolog or record_id that it could be trying to plot? It seems nonsensical that the error says length should be 59 and you are saying that there are 59 values. Perhaps, if it one variable is a factor it could have vestigial levels. — Tanner33, Sep 01 '20 at 20:23
@chemdork123, just tried to do that, but the same problem remains. — Lucas Lazari, Sep 01 '20 at 20:30
... as the error is related to the `fill` aes (see the error message) I would try making `fill=value` a local aes to `geom_bar`. — stefan, Sep 01 '20 at 20:32
Yep just found that it's in the error message and was going to post something until I saw this was answered. — chemdork123, Sep 01 '20 at 20:32

DaveArmstrong · Accepted Answer · 2020-09-02T18:54:08.517

Without access to the data you'll have to confirm, but if you remove the data and aesthetics from ggplot() and put them in geom_bar(), it should work:

ggplot() + 
    geom_bar(data=t, aes(y=as.factor(record_id), x=Values, fill=value), 
        position=position_stack(reverse= TRUE), stat="identity") +
    geom_point(data = new_df, aes(x=sorolog, y = record_id), 
                colour = "#a81802", size = 4, shape = 1)

EDIT

I am adding an answer to the question about adding a color legend for the points. Also added size and shape to the points, too.

ggplot() + 
  geom_bar(data=t, aes(y=as.factor(record_id), x=Values, fill=value), 
           position=position_stack(reverse= FALSE), stat="identity", width = 0.5) +
  scale_fill_manual(values = c("brown1","chocolate1"),name = "", 
                    labels = c("Hospitalization to Discharge", "Symptom to Hospitalization")) +
  geom_point(data = new_df, aes(x=sorolog, y = as.factor(record_id), colour="Point Label 1",
                                size="Point Label 1", shape="Point Label 1")) +  
  geom_point(data = new_df, aes(x=final, y = as.factor(record_id), colour="Point Label 2", 
                                size="Point Label 2", shape="Point Label 2")) + 
  scale_colour_manual("points", values=c("Point Label 1" = "darkcyan", "Point Label 2" = "darkred"), 
                      labels= c("Point Label 1", "Point Label 2")) + 
  scale_shape_manual("points", values=c("Point Label 1" = 1, "Point Label 2" = 16), 
                      labels= c("Point Label 1", "Point Label 2")) + 
  scale_size_manual("points", values=c("Point Label 1" = 5, "Point Label 2" = 4), 
                     labels= c("Point Label 1", "Point Label 2")) + 
  theme_minimal()+
  labs(title="Patient timeline - from symptoms to hospitalization and discharge",
       x ="Days", y = "Patient ID")+
  theme(text = element_text(family = "Garamond", color = "grey20"))

The trick here is to put all of the point attributes - colour, size and shape, in the aesthetics with the same labels. The attributes themselves supplied to values need to be named vectors where the names are the same as the aesthetic names. I found this post helpful in putting the pieces together.

The main idea is that you have to add a colour aesthetic to the points, but it doesn't have to come from a variable in the data frame, you can make it up on the fly.

It worked, I just needed to put y=as.factor(record_id) in the geom_point(). Thank you — Lucas Lazari, Sep 01 '20 at 20:32
do you have any idea on how to add two different legends, one for the geom_bar and the other one for the geom_point? — Lucas Lazari, Sep 02 '20 at 16:59
@LucasLazari I added the answer to the answer for the original question — DaveArmstrong, Sep 02 '20 at 17:06
It worked, however the two different points stacks, I need them separated.. I'm trying to figure it out... I edited the post with the result — Lucas Lazari, Sep 02 '20 at 17:44
@LucasLazari what do you mean by separated? Do you mean you want them to be next to each other rather than on top of each other? If so, then you can use `position = position_nudge(y=.2)` for one set of points and `position = position_nudge(y=-.2)` for the other set of points — DaveArmstrong, Sep 02 '20 at 18:27
For example, "Point Label 1" should be only the empty blue circle. And the "Point Label 2" should be only the filled red circle. In the legend both are stacked together. — Lucas Lazari, Sep 02 '20 at 18:30
@LucasLazari I re-edited the answer to include the size and shape aesthetics. This should get you where you need to go. — DaveArmstrong, Sep 02 '20 at 18:54
It worked perfectly, thank you very much! I was almost getting there.. I used the melt function to the new_df, that simplified a lot removing one geom_point(). But I was still struggling yo understand the aesthetic part. Again, thank you! — Lucas Lazari, Sep 02 '20 at 18:57

Stacked bar plot with points, but with different aestetics length - ggplot2

1 Answers1