# how to present 4 variables with binary outcomes?

So I have the following data.

Index , Years , Negative , Positive , Category
1 , 2004 , 23 , 29 , A
2 , 2005 , 23 , 29 , A
3 , 2006 , 30 , 22 , A
4 , 2007 , 21 , 31 , A
5 , 2008 , 19 , 33 , A
6 , 2009 , 24 , 28 , A
7 , 2010 , 35 , 18 , A
8 , 2011 , 26 , 26 , A
9 , 2012 , 30 , 22 , A
10 , 2013 , 23 , 29 , A
11 , 2014 , 22 , 30 , A
12 , 2015 , 27 , 26 , A
13 , 2004 , 26 , 26 , B
14 , 2005 , 30 , 22 , B
15 , 2006 , 26 , 26 , B
16 , 2007 , 28 , 24 , B
17 , 2008 , 20 , 32 , B
18 , 2009 , 32 , 20 , B
19 , 2010 , 31 , 22 , B
20 , 2011 , 31 , 21 , B
21 , 2012 , 31 , 21 , B
22 , 2013 , 22 , 30 , B
23 , 2014 , 27 , 25 , B
24 , 2015 , 21 , 32 , B
25 , 2004 , 19 , 33 , C
26 , 2005 , 36 , 16 , C
27 , 2006 , 23 , 29 , C
28 , 2007 , 18 , 34 , C
29 , 2008 , 27 , 25 , C
30 , 2009 , 24 , 28 , C
31 , 2010 , 29 , 24 , C
32 , 2011 , 27 , 25 , C
33 , 2012 , 24 , 28 , C
34 , 2013 , 26 , 26 , C
35 , 2014 , 32 , 20 , C
36 , 2015 , 30 , 23 , C

And I'd like to present the data. I've tried using 3d barplot but the result isn't visually pleasing.

• (1) Where are the "binary outcomes" of the title? None of these variables is binary. (2) What is the purpose of your data presentation? Who is the audience? What information is the presentation intended to convey? What will be the medium? (3) What do the data mean? Is "Years" a date? Is there some determined relationship between "Negative" and "Positive" (they seem to sum to 52 or 53 in all cases)? What other special structure might your data have?
– whuber
May 22 '17 at 21:34
• Ah I forgot to add that negative and positive are the counts for the binary outcomes May 23 '17 at 6:50

You described the problem as one of presenting four variables, but since,
Negative = 52 - Positive except for two years with 53 weeks. I would say that there is little gained by presenting both Positive and Negative. I will look at Year, Positives and Category.

You say that you want to present the data. I guess that how to do this depends on what "present" means. If you just want people to be able to see the values, a table is likely your best bet.

                              Positives
2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015
A   29   29   22   31   33   28   18   26   22   29   30   26
B   26   22   26   24   32   20   22   21   21   30   25   32
C   33   16   29   34   25   28   24   25   28   26   20   23


I suspect that was not your intent, but in order to do better, you need to figure out what it is you want to say with your visualization. "The purpose of visualization is insight, not pictures." In order to figure out what you want to present, you probably need to do some exploration of the data and discover something interesting to say.

You could imagine that there could be something interesting in viewing the data as a time series - some possible change in the number of positives over time. The straightforward way to plot the data produces a messy jumble that does not seem to show much.

Separating the Categories into separate line charts makes it much more readable, but does not seem to reveal anything about the data over time.

Let us imagine then, that each Category just exhibits random fluctuations about a mean. Are the three categories the same? Plotting the density of each category starts to show something more noteworthy. The three Categories seem to have different means.

Category C seems to be well centered and have greater variance than the other two categories. Category A has a higher mean and skews left. Category B has a lower mean and skews right. I am not sure if that means anything in your problem space, but it seems interesting to me.

There may be additional information in here, but you can't expect arbitrary graphs to be interesting. You must discover something about the data and then make graphs that illustrate the interesting relationships.