Discussion


Note: Images may be inserted into your messages by uploading a file attachment (see "Manage Attachments"). Even though it doesn't appear when previewed, the image will appear at the end of your message once it is posted.
Register Latest Topics
 
 
 


Reply
  Author   Comment  
RodA

Registered:
Posts: 3
Reply with quote  #1 
Hi,
I'm seeking the opinion of the good people of this board. This chart shows the percentage of Israeli-Arab participation in the general elections in Israel. It's obviously horrible - 77 percent is above 79 - but ignoring that, I want to use this line graph as an example of how not to do things in a lecture I'm going to give. 
I think a line graph is dead wrong here, because the X axis doesn't have set intervals. The time between general elections in Israel is 4 years, but seldom do governments see their period through - So 1992-1996 spans 4 years, but 1996-1999 spans three, and 1999-2001 spans but two years. Still. they are located at equal distances from one another on the axis, which I think distorts the data and shows trends as steeper than they really are. 
On the other hand, I wouldn't use an interval year axis anyway, because the change is between elections, not years. 
I thought I'd use a bar chart with a bar for every election, but a bar chart would necessitate starting at zero, which would obscure some of the subtle differences.
So I'm thinking maybe a dot chart? Is it alright to use a dot chart and have the Y axis not start at zero? I'd love to hear your opinions.
arab_hazbaa.jpg

danz

Registered:
Posts: 186
Reply with quote  #2 
I would do this:

Build a scatter chart with correct values Percentage(Year) and connect consecutive (in time) points using a thin line.
Keep for axis labels 1950, 1960, 1970, 1980, 1990, 2000, 2010 and use thin lines for grid.
Keep the values scale 0-100%. It is all-right considering the variation of data between 18% and 90%.
I do not use labels in line charts. If is a must, then label only some relevant values: 79% in 1949 (first), 90% in 1955 (max), 68% in 1981 (local min), 77% in 1996 (local max), 18% in 2001 (min), 53.4% in 2009 (last). Use markers only for the labeled values.

A short text below the chart describing the trend would be beneficial: "In 1949, when were organized first elections, 79% of electors were Israeli people. In 1955 the percentage went to 90% (maximum). After that, the percentage continue to decrease to 68% in 1981. After a slight ascendant trend till 1996 when percentage went to 77%, the percentage resumed the descendant trend reaching 53.4% in 2009, with special situation (...) in 2001 when percentage went to just 18%".

If all the values are important, add somewhere else in the view or document a separate table.

In short you will see the trend, a few relevant values in a correct scaled graph and the interpretation of the graph. 

jlbriggs

Registered:
Posts: 194
Reply with quote  #3 
I agree with danz that an axis running from 0-100 with this data set would not be a problem - the data spans most of that range already. On the other hand, I would probably not plot this as a bar chart anyway.

I am not sure that I see any difference between a line chart and a "scatter plot with the points connected by a line", however. They are the same thing in this instance (perhaps you meant that a line chart might not account for proper x axis scaling? no reason that it can't though...)


I do think that a line chart is still the most appropriate way to show this data. Even though you are talking about the change from election to election, those elections have sequential dates, and you are plotting change over time.
The differences in years between elections do not vary so much that your scale would be affected in a negative way, IMO.
I would, absolutely, set the x axis points at the appropriate, scaled by year, distance from each other.

I think the bigger issues here are

1) making sure the data is properly plotted (as you mention, 77 can't be higher than 79...),

2) that the labels, grid lines, and other visual elements are reduced to only what they need to be (axis labels used sparingly, data point labels removed unless a point of particular note is to be highlighted, grid lines thin and sparse, background image removed, etc)
danz

Registered:
Posts: 186
Reply with quote  #4 
jlbriggs is right about not being any difference between a line chart and the solution I mentioned. However software like Excel implements a line chart as a categorical chart, but a scatter plot with connected dots allows the right time scale. 
RodA

Registered:
Posts: 3
Reply with quote  #5 
Hi,
I appreciate your input and taking the time to answer. I agree on some of the things, such as danz's claim that we can use a 0-100 y-axis, but I respectfully disagree a line chart is preferable here. 
I think jlbriggs got my point regarding the x-axis just right: If we leave the x-axis as it is, we risk showing trends which are steeper than what they really are over time. But the more I think of it, I don't think this is a time-series chart per se. Spreading the data across a yearly axis isn't "right" in the sense that we don't get yearly data. The change, one might claim, isn't really over time as it is over elections. Elections have many other variables other than the year they take place in that may affect the data: Who's the incumbent? Who are the contenders? What are the issues? For example, the 1973 elections show a relatively big decrease in Israeli-Arab participation - partially because the elections were held after the 1973 war. A better example are the 2001 elections, which were the only time (And the law has been fixed so this never happens again) that elections were held only for prime minister and not for the parliament. So, it seems to me that there are many attributes which are intrinsic to elections, that warrant showing each election as it stands on its own. 
A bar chart also allows me to add some more data, specifically - to stress out the years where there has been change of government, thus showing the relation, or there lack of, between Israeli Arabs participation and government overturn. With the use of more colours I could have also  shown in what direction did the government change (Right to left or left to right), but I decided against it as it seemed too...burdening. 
Here's my mock up - I'd love to hear your honest criticism of it. 


Arab Participation.png 

jlbriggs

Registered:
Posts: 194
Reply with quote  #6 
@danz - I thought that might be what you were referring to, but I haven't made a chart in excel for so long...  ;)

@roda - I understand your points in regard to the various factors surrounding an election, but I don't agree at all that it has any bearing on which chart to use.

Additional information can very easily be added to either a bar or line chart, and I would argue that a line chart makes it easy to offer such highlighting, as you don't have so much space taken up by each data point - you can change the color of the background at certain points of the x axis, change the color of a line point marker, and arrows, lines, flags, labels, etc...

And I still disagree with your assessment of whether this is time series data - the time element is absolutely an important one in regard to this data, and the elections you are plotting occur in sequential time progression, at semi-regular intervals.  Showing the difference in the timeframe betwteen sequential elections seems to me like one of those extra data points that would be relevant to show.




danz

Registered:
Posts: 186
Reply with quote  #7 
Roda,

I see no reason to use a bar chart instead of a line chart for the presented case. You can use markers for special situations and separate charts/views for the other factors (like population, type of elections statistics, etc). A simple chart as bellow should be OK. It contains four markers, two of them having distinct colors.

elections.png 



jlbriggs

Registered:
Posts: 194
Reply with quote  #8 

@danz - fully agree. I made a similar chart this morning, though after I made mine, I saw yours and updated my data point label text because I liked yours better;)

But, for reference - my version:

chart(6).png 
My version uses the actual election years as the x axis tickmark points, which I did just as an example - I am on the fence as to whether I prefer that, or a standard interval.  But I want to show it as an option to make the difference in years between elections more visually evident.

In addition, there appear to be 6 points highlighted in the bar chart example - annotations could easily be added for them as well, including the left/right shift mentioned earlier.

Or, those elections could possibly be broken out into a second chart that had more detail, which would leave the overall view clean, while providing richer context as well.


danz

Registered:
Posts: 186
Reply with quote  #9 
@jlbriggs - I think I will stay on my side of fence ;)

I do not see how relevant is the "irregular intervals" information for a graph which keeps the descendant trend independent of elections intervals.
I can hardly notice in graph the intervals variation, I am sure that just having a comment about it is enough. 

Initially I added some ticks/markers on time axis visually suggesting the irregular intervals for elections. I didn't like them, so I removed them.

I am not a big fun of complicated graphs, I added just a few markers for "Elections overturned".  

Anyway, here it is my adjusted version.
 elections2.png 
 
 

jlbriggs

Registered:
Posts: 194
Reply with quote  #10 
@danz - yeah, I fully get it. I am not necessarily advocating for the way that I set up the x axis, just wanted to put it out there as an option, and I think it can have some relevance, and if it has relevance, showing it visually in the axis labeling is a pretty easy and straightforward way to do it.

Previous Topic | Next Topic
Print
Reply

Quick Navigation: