Friday, 19 April 2013

Causal correlation or casual correlation

The following story is included in its entirety.  Leaving aside all the effort which went into collecting the data, what does it say?  

It doesn't say that if you asked the local council to rename your street, then your house price would automatically change.  All that it is saying is that there is an interesting relationship between street names and house prices.  The examples given of Upper Phillimore Gardens and Zeus Lane illustrate this;  exchanging the names would have little effect on the prices - because they depend on the location, size, facilities, etc of the properties on each road.

Added to which, I have commented elsewhere on the methods used by estate agents to establish the current value of houses.  Look at the precision of the figures here "the average home is worth £5,740,496."  There are 53 properties in that road, and the algorithm for estimating their value gives a figure correct to the last pound.  Now the properties vary from detached through semi-detached to flats.  So the average is the average of dissimilar properties.  (Adding grapes to oranges and melons to give the average for fruit.)  Presumably their average is the mean.  But is that an appropriate statistic to use, for data which are skewed, and by definition non-negative?

The house that I grew up in was in a road whose name began with R (£223,241).  Most of my postgraduate years I lived in a road starting with S (£228,237)  In Exeter, streets for my three homes have started with F, W and M (£228,489, £231,145, £221,498 respectively)

However, in the same order, those five streets have the following current average property prices:
£403,716  (A village with large houses)
£197,501
£190,002  (Exeter when I was single, where the properties included flats)
£217,555  (Getting married moved me up in status?)
£595,308

So what?  As this blog is about Operational Research, I suppose that the story reminds me that one must ask whether we always ask the right critical questions about our analysis.

Vowel play – street names starting with U have highest av. property values #zooplafacts

UpperPhillimoreGardens
Forget streets with the X-factor, living on a street with a name beginning with the letter ‘U’ could mean your property is worth a lot more, according to Zoopla.co.uk research.
We analysed average house prices on more than three-quarters of a million streets in the UK. The most expensive street beginning with ‘U’ is Upper Phillimore Gardens where the average home is worth £5,740,496.
Average property prices on streets that start with a ‘U’ currently stand at £251,307 – the highest of any letter in the alphabet – and £25,503 more than the current average UK property value of £225,804.
At the other end of the scale, streets beginning with the letter ‘Z’ have the lowest property values, with the average property worth just £180,046 – 20% less than the UK average. One particular example is Zeus Lane in Waterlooville, Hampshire. Despite being named after a Greek god, property values on the street are far from Olympian. The average property value on the street is just £141,278, 37% lower than the national average of £225,804.
Streets beginning with the letters T (£248,008) and O (£244,450) came in second and third respectively in terms of the highest average property values, while those starting with the letters J (£190,802) and A (£211,744) record the second and third lowest property prices.
And vowels beat consonants hands down when it comes to house prices. The average property on a street beginning with a vowel is worth £6,306 (3%) more than a street beginning with a consonant. Street names beginning with a consonant have an average property value of £222,789, compared to £229,095 for those with a vowel.
So, it seems that there may be more to a street name than meets the eye. With a £70,000 difference between the average property price on streets starting with the letters U and Z, and a £6,000 difference on average on streets beginning with vowels and consonants, it is advisable to select your street wisely!

The A-Z (or U-Z!) of property values

Starting Letter No. of UK Streets Avg. Property Value
U 3,565 £251,307
T 44,927 £248,008
O 16,358 £244,450
H 53,223 £234,045
C 77,769 £233,873
P 36,065 £233,550
W 50,979 £231,145
F 25,835 £228,489
L 39,563 £228,400
S 71,304 £228,237
B 71,884 £227,105
Q 3,826 £226,824
N 18,475 £225,911
R 31,154 £223,241
M 51,216 £221,498
K 16,728 £221,106
G 33,335 £219,248
I 4,316 £219,125
E 19,216 £218,849
D 25,395 £218,643
Y 2,751 £218,367
V 7,752 £217,240
A 31,333 £211,744
J 5,565 £190,802
Z 180 £180,046
As always, please feel free to share and use this info, all we ask is that you credit the source as Zoopla.co.uk and link to Zoopla.co.uk. Thank you.

The joy of ... academic papers?

Earlier this year, I wrote about Father Geoffrey's journey (here), the account of a piece of research which I had written up for publication.  It is now freely available (OR Insight 26: 140-148, doi:10.1057/ori.2012.16, here or pdf).  Yesterday, copies of the journal issue arrived to subscribers.  This morning, there was an email from one of my oldest friends and colleagues in Operational Research, reading:

Dear David,
I have a confession.  I enjoyed your article.

I suspect that some of that email was written, tongue-in-cheek, to let me know that he had seen it and had read it.  And he was doubtless aware of the old joke that only three people ever read research papers - the author and the two referees.  So it is pleasant to know that at least one person read the paper on the day it appeared. 

However, there is a deeper aspect to the message.  How often do we enjoy an article in our research journals?  I mean research journals, not our professional journals, which should have a lighter style.  Put you hand on your heart and ask "Which papers over my O.R. career have I enjoyed reading?"  For many of us, the answer will be embarrassingly few.  I can think of half a dozen, in each of which the personality of the authors came through, as they wrote about their enthusiasm for the work that they were describing.  And these papers still maintained academic rigour. 

The academic style has been lampooned numerous times; I really enjoyed the spoof reports of an 11 month old girl and her toys (description and "academic" paper) as an example.

How could your next paper or presentation be prepared so that people might "Enjoy" it?


Tuesday, 9 April 2013

Models for thinking with - right or wrong?

The first encounter that I ever had with Operational Research was an article with the title "Models for thinking with".  It was a very simple introduction to the science of O.R.  Despite its simplicity, it brought out some of the features of an O.R. study, especially the idea of using a model to represent the underlying reality of a business situation or problem.

And I suppose that one of the characteristics of good O.R. is the ability to look at "the problem" and devise ways of modelling it, to include data and decisions.  In other contexts, one might call it analysis, but that has other meanings to mathematicians and computer scientists, so I use the term with caution.

There are times when there are several ways of constructing a model.  I forget which leading O.R. academic, back in the late 1940s or early 1950s, attended a lecture on the simplex method for solving linear programming models.  He remarked afterwards that he doubted if the technique would be of great value, because "the world is nonlinear".  His models of the world reflected that nonlinearity; linear programming models give a different model of the world.  Which is correct?  There are other lessons from history where there have been two (or more) models of a real world situation, and with the passage of time, one has been shown to be incorrect or inadequate.  There was a computer engineer who predicted that the US would need four or five computers, back in the 1950s.  (So now we have three computers and a smartphone in our study.)

I have blogged in the past about when two models come into conflict, and from time to time, I come across further examples.  So here are a few from Devon and Exeter, though most can be replicated in many other places.

(1) How far apart should houses be built?  Many semi-detached houses in the city, built in the 1930s to 1950s, were constructed with a shared drive between the houses, leading to a Y-fork and two garages.  The width of the shared drive was designed for the cars of that era, and modern cars are too wide to fit.  The architect's model of the need was based on current observations (of the size of cars), and did not consider the consequences of any change in the data.  Surprisingly, older houses in the city are better equipped for garages, as they were built with a service road at the rear of the property which can give access to cars.

(2) What places should be on signposts?  Most signs are put up by people who are local and know the area.  Their models of the needs reflect their knowledge.  On the other hand, those who are strangers to the area need different information.  Sometimes the two are in conflict.  A few years ago, a lot of signs went up around here, pointing to "Torbay", as a generic term, known to locals (and thus part of their model of the area) for Torquay, Paignton and Brixham.  When it was realised that people heading for Torquay did not identify Torbay with Torquay, the signs were changed.  Within the city, one of the car parks has been renamed "John Lewis car park" and there is some confusion as the name does not appear on older street maps, and has only just started to feature on satnavs.  The model for signage is a local one, not one aimed at the general user. 

(3) Following on from (2) are the signs on long distance routes.  There are many examples of signs which indicate a distant town, and the next sign that the traveller passes on the way there has no mention of that distant town.  The signs mayhave been designed and erected by two people with different models of the need of the motorist.

(4) The book and film and play "War Horse" (written by Devon author Michael Morpurgo) draws attention to two models of warfare that existed at the start of the first World War.  Some army staff believed in the persistence of cavalry, but the cavalry charge was defeated by a model of warfare that used machine guns. 

So, lessons -- what assumptions are you making about your models?  Are they reasonable?