An American named Zinsser described writing thus: "the difference
between the right and the wrong word is like the difference between
lightning and a lightning bug." I've struggled to find a UK equivalent
of lightning bug, which makes his point.
When teaching graphics, I struggle to get students to (a) see what is
there rather than what they expected or thought was there (b) experiment
with a variety of graphs and see that each delivers a different message,
and (c) hence choose the graph that conveys a message they intend and to
justify their design choices.
It starts with software marketed under the slogan "use our software and
*with a few clicks* you'll create professional quality graphs." Can you
imagine "use our word processor and with no effort you'll write
bestselling novels"?
Having mentioned data density, Significance (June 2010 p69) has
wonderful examples of two bar charts occupying 12 column inches (>1/3 of
the page). Each chart has just two bars that happen to total to 100%,
so each is effectively displaying just one number. The authors may have
wanted to draw attention to the four factor levels that contributed zero
to the sum but that's not explicit in the graph or the text - so relies
on readers seeing what's *not there* and inferring. I would have
expected my graphics students to suggest stacked bars (using the Y scale
to 100) and merging Figures 1 & 2 (& 3?) to put the bars side by side to
ease comparison of the child/family/secondary components. Use
colour/pattern/tone matching to visually link the components not to
spuriously distinguish the three figures. I'm not saying they should,
just indicating a trail of design thought. BTW, do you find the
"credible intervals" credible as statements of the potential
variability?
No offence taken, but I replied off-list to Julian and had not intended
to indulge myself during work hours.
Allan
-----Original Message-----
From: Wells, Julian [mailto:[log in to unmask]]
Sent: 17 June 2010 10:02
To: Allan Reese (Cefas); [log in to unmask]
Subject: RE: How to lie with statistics [was: The state of the nation]
I certainly would deny their right to choose to describe a one per cent
increase in something as "rocketing", or a 100 per cent increase as
"slight".
Whether I would want to *prevent* such abuses, given the opportunity, is
another question: if I was refereeing a submission to a scholarly
journal I'd certainly have something pointed to say in my report, at the
very least.
Julian
________________________________________
From: Allan Reese (Cefas) [[log in to unmask]]
Sent: 17 June 2010 09:34
To: Wells, Julian
Subject: RE: How to lie with statistics [was: The state of the nation]
You might disagree with their choice but I hope you would not question
(ie deny) the author's right to choose.
Or am I just out of touch with modern education? I do recall stories of
exam papers being downgraded because
examiners were too stupid to recognise synonyms of model answers. ;-)
Regards
Allan
-----Original Message-----
From: email list for Radical Statistics [mailto:[log in to unmask]]
On Behalf Of Wells, Julian
Sent: 17 June 2010 07:10
To: [log in to unmask]
Subject: Re: How to lie with statistics [was: The state of the nation]
Allen writes:
'You would not question the right of authors to choose words in the
text, eg between "has crept up", "has increased", "has rocketed".'
Well, I would -- if the numerical increase was, say, 1 per cent rather
than 100 per cent (or v.v.).
That facts only have meaning in the light of context and interpretation
does not give one the right to describe them in any old way one likes
...
Julian
Dr Julian Wells
Acting Director of Studies
School of Economics
staff web-page: http://fass.kingston.ac.uk/staff/cv.php?staffnum=287
personal web-site: http://staffnet.kingston.ac.uk/~ku32530
Senior lecturer in economics
School of Economics
Faculty of Arts and Social Sciences
Kingston University
Penrhyn Road
Kingston-upon-Thames
KT1 2EE
United Kingdom
+44 (0)20 8417 2285
________________________________________
From: email list for Radical Statistics [[log in to unmask]] On
Behalf Of [log in to unmask] [[log in to unmask]]
Sent: 16 June 2010 23:37
To: [log in to unmask]
Subject: Re: [RADSTATS] How to lie with statistics [was: The state of
the nation
Scaling pictures linearly with data is a classic error. It raises the
question of whether specific software was used, in which case blame the
software writer, or the figures were given to a graphic artist who was
just innumerate. I think there are psychology studies (but don't have
refs to hand) showing that viewers compare areas in this situation but
not exactly. Square root scaling would therefore be better than linear.
Had they used pictograms (silhouette of wheelchair), readers would be
psychologically comparing volumes and a cube root transformation would
be advisable. Making the pies vary in colour and tone exaggerates the
difference, particularly between 1997 and 2003.
Note that three discs are used to represent three data items, so the
data density (cf Tufte) is very low. This is not necessarily bad; you
might deliberately include such a graph to lighten a text-heavy page.
Bars would more naturally match the values, and the spacing between bars
could scale with the number of years.
The question for the authors is what message they intended the graphic
to convey. If it is "alarming growth", they may be justified in adding
"stress" rather than "distortion" - the assumption that these figures
are "correct" and must therefore be represented with precision is in
actuality an assumption. You would not question the right of authors to
choose words in the text, eg between "has crept up", "has increased",
"has rocketed". My worry with graphics is always that the authors (and
proof-reading editors - cf the Radstats comment that two parts of
thereport are inconsistent) are not sensitive to the implicit messages
in graphs. Too many graphs in publications look like exam scripts - you
guess what they're getting at and give credit for effort!
Allan
-----Original Message-----
From: email list for Radical Statistics [mailto:[log in to unmask]]
On Behalf Of Ted Harding
Sent: 16 June 2010 10:40
To: [log in to unmask]
Subject: Re: How to lie with statistics [was: The state of the
On 16-Jun-10 08:02:04, Wells, Julian wrote:
> Yikes! The Cabinet Office people might profitably be directed to the
> ONS house style guide or its predecessor "Plain Figures", not mention
> that old faithful "How to Lie with Statistics" (I see from Wikipedia
> that the latter has been claimed to have sold more copies than any
> other text on statistics).
>
> Julian
I was so struck, looking at the graphic, by the apparent disparity
between the perceived "sizes" of the circles and the associated
numbers that I even wondered if the sizes had been deliberately
fudged! So I did a check. Enlarging the picture and using a ruler,
I got, for the diameters of the circles:
1.2m: 42mm
1.5m: 53mm
1.8m: 64mm
Now:
1.5/1.2 = 1.25 ; 53/42 = 1.26
1.8/1.2 = 1.50 ; 64/42 = 1.52
1.8/1.5 = 1.20 ; 64/53 = 1.21
So pretty close agreement between diameter and value (given that
the edges of the circles were a bit fuzzy anyway).
Which just goes to show (essentially Julian's point) that the eye
is more influenced by relative areas than by relative linear
dimensions. Indeed, attempting to estimate "by eye" the relative
diameters (before using the ruler) I had judged that successive
circles were more like 1.5 times as wide as their predecessors.
And that is in keeping with ratios of areas (1.2^2 = 1.44). A nice
"optical illusion".
Well, now, the Cabinet Office statisticians should of course be
encouraged to make the areas in the right ratios, rather than
the diameters, in order to induce appropriate perceptions.
And then they would have to make diameters (presumably the defining
quantity for producing the graphic) proportional to the square roots
(which might require a coffee break). And then they would have to
explain that somehow. Or not -- the report makes no attempt to explain
the scaling of the circles.
So maybe they wouldn't explain. Then some journalist could do what
I've just done and then report "The Cabinet Office report includes
graphics that deliberately play down the massive increase in
claimants from 1997 to 2009. The increase from 1.2m to 1.8m is
an increase of 50 per cent. The circles increase in size by just
[journalist's synonym for 'only'] 22 per cent."
But what journalist is going to pick up on the apparent exaggeration
that is present in the graphic as presented? Maybe Tim Harford?
So perhaps the C.O. should explain! A simple parenthesis in the
caption (to the better form of the graphic) would do it:
"Figure 3.1: The numbers of working-age Disability Living
Allowance claimants have increased by over 40% since 1997,
from 1.2 million to 1.8 million (areas of circles proportional
to numbers)."
Or is that too abstruse?
But since they've stated the numbers so clearly, why bother with
the graphics?
Well, you sort of can't help it once you've woken up a spreadsheet!
In the memorable words of one of Jasper Carrott's best:
"Q: Why do dogs lick their ****s? A: Because they can."
I think we have a very long way to go before the public understanding
of statistical information is informed by presentations which induce
correct perceptions.
Ted.
--------------------------------------------------------------------
E-Mail: (Ted Harding) <[log in to unmask]>
Fax-to-email: +44 (0)870 094 0861
Date: 16-Jun-10 Time: 10:39:23
------------------------------ XFMail ------------------------------
******************************************************
Please note that if you press the 'Reply' button your
message will go only to the sender of this message.
If you want to reply to the whole list, use your mailer's
'Reply-to-All' button to send your message automatically
to [log in to unmask]
Disclaimer: The messages sent to this list are the views of the sender
and cannot be assumed to be representative of the range of views held by
subscribers to the Radical Statistics Group. To find out more about
Radical Statistics and its aims and activities and read current and past
issues of our newsletter you are invited to visit our web site
www.radstats.org.uk.
*******************************************************
************************************************************************
***********
This email and any attachments are intended for the named recipient
only. Its unauthorised use, distribution, disclosure, storage or
copying is not permitted. If you have received it in error, please
destroy all copies and notify the sender. In messages of a non-business
nature, the views and opinions expressed are the author's own and do not
necessarily reflect those of the organisation from which it is sent.
All emails may be subject to monitoring.
************************************************************************
***********
This email has been scanned for all viruses by the MessageLabs Email
Security System.
****************************************************** Please note that
if you press the 'Reply' button your message will go only to the sender
of this message. If you want to reply to the whole list, use your
mailer's 'Reply-to-All' button to send your message automatically to
[log in to unmask] Disclaimer: The messages sent to this list are
the views of the sender and cannot be assumed to be representative of
the range of views held by subscribers to the Radical Statistics Group.
To find out more about Radical Statistics and its aims and activities
and read current and past issues of our newsletter you are invited to
visit our web site www.radstats.org.uk.
*******************************************************
This email has been scanned for all viruses by the MessageLabs Email
Security System.
******************************************************
Please note that if you press the 'Reply' button your
message will go only to the sender of this message.
If you want to reply to the whole list, use your mailer's
'Reply-to-All' button to send your message automatically
to [log in to unmask]
Disclaimer: The messages sent to this list are the views of the sender
and cannot be assumed to be representative of the range of views held by
subscribers to the Radical Statistics Group. To find out more about
Radical Statistics and its aims and activities and read current and past
issues of our newsletter you are invited to visit our web site
www.radstats.org.uk.
*******************************************************
This email has been scanned for all viruses by the MessageLabs Email
Security System.
This email has been scanned for all viruses by the MessageLabs Email
Security System.
******************************************************
Please note that if you press the 'Reply' button your
message will go only to the sender of this message.
If you want to reply to the whole list, use your mailer's
'Reply-to-All' button to send your message automatically
to [log in to unmask]
Disclaimer: The messages sent to this list are the views of the sender and cannot be assumed to be representative of the range of views held by subscribers to the Radical Statistics Group. To find out more about Radical Statistics and its aims and activities and read current and past issues of our newsletter you are invited to visit our web site www.radstats.org.uk.
*******************************************************
|
