How not to do it: Graphs that don't mean anything

Take a look at the following graph (click on image for larger version). It shows the number of long-term (6 month) detentions under the Mental Health in Scotland between 1985 and 2012:

Original here.

You can see a few problems:

The axes aren't labelled (although it should be fairly obvious)
The data aren't labelled - are they plotting one or two sets of data?
The years are a mixture of single years and ranges. When did years start and end, exactly?
In the year 'year', there were no detentions. However, we're pretty sure that there wasn't a year 'year' and this is meant to be the axis title (whoops).

I want the graphs to look flat

This is how we would have displayed the data:

Same data, different effect. All we have done is plotted natural logs of the data, perhaps in order to smooth out variation. We've also made sure that the y-axis values are quite wide, making the line look flatter. We might reasonably conclude that there has been little change in detention rates between 1985 and 2012 (looking at the graph).

After all, who really knows what natural logs are about and we can always argue that we are reducing the residuals, anyway.

I want to make the graphs look very jaggedy

Simple. We just choose our y-axis range to pick the narrowest possible (between 300 and 1300). This makes the differences look greater.

I want the graphs to make it look like I did something clever

If you wanted to show that something you had done in 2005 had changed the number of detentions in Scotland, you would choose the above graph and modify it a bit. You might be applauded for changing a worrying trend, perhaps.

If you ignore the grey line (the original data) and focus on the trendline (black, dotted line), you can see that rates of detention started coming down after 2004-05 and you might legitimately claim that they are lower than they have been for 5 years!

If you're going to do this, it would make sense not to mention that the Mental Health Act changed in October 2005, and changes in practice brought about by the new Act almost certainly explain any change seen.

It probably wouldn't be wise to mention that numbers are going up steadily since 2009, or that the numbers in 2012 are about three times what they were in 1985 (but hey, you're not showing these data).

Conclusions

The underlying values are the same, yet on one graph we can conclude that detention rates haven't changed, and another they appeared to be increasing at a tremendous rate.

Which graph you choose to use depends on you, the customer. We can help you to make that choice.

Disclaimer: We use these graphs and data to illustrate some examples. MGGD Consultancy does not claim to have any association with organisations mentioned.