News Numbers - Helipcopters

With plenty of numbers being used in the news, it's often very tempting to overplay the effect of the numbers, often quoting without some context.

Hence having personally coined the term "Numberator"....the use of a number without some necessary contextual denominator here's another example....Another of those news number naughties.... or news numberation.

So the UK is very helpfully sending aid to the Philippines to support victims of typhoon Haiyan.  HMS Darling is helping out but will be replaced by HMS Illustrious announces the Prime Minister...

"And I can announce today that once Daring has started its work, we are actually going to be able to replace in time HMS Daring with HMS Illustrious which is of course a carrier with helicopters, seven times as many helicopters as on HMS Daring and with the key ability to process fresh water. So we will be giving further assistance in the best way we can". 

So "seven times" as many helicopters, that's quite some emphasis.  That's also the edited sound bite that made the main news headlines.  So how many would that be in practice?

Actually it's seven helicopters in total.  HMS Darling had one and HMS Illustrious has seven.  So yes technically seven more.

"Seven times more helicopters" does seem to have a more powerful emphasis than "Seven helicopters". Seven times is a relative statement that can apply in multiple situations for higher numbers.  So if there were 2 helicopters on HMS Darling there could be 14 on HMS Illustrious, if 3 helicopters on Darling then 21 on Illustrious.    So while the relative statement can be applied for larger (and even much larger) scenarios, here it's applied in the lowest possible extreme scenario, but at the same time can, and arguably does, imply larger scenarios.  In the larger scenarios the "7 times" is helpful shorthand.  Rather than saying something in increased from 123 to 861, the "7 times" gives a better quicker sense.   So when we hear "7 times more" were more naturally drawn to assume it's shorthand for bigger numbers than simply 1 to 7.  

An there's another twist.  Of course while HMS Illustrious does have seven times more helicopters than the one on HMS Darling which is already deployed. The net increase in the Philliplines is of course only 6 extra helicopters.

One simple calculation step away from the raw data, but looses so much meaningful context in the presentation.  It's also a comparative measure rather than an absolute one.

This is a great simple demonstration of the powerful message around exploratory data analysis  The further we are from the underlying data, the greater the chances of misunderstanding and misrepresentation. Add to that the messages from the Statistical Process Control movement,  "No data has meaning apart from it's context".

Statistics Views - Summit

For a couple of days, the great and good of planet earth's statistics academics and closely associated players
came together to take stock on the past, present and future.  All as part of the International Year of Statistics, and hosted by our own Royal Statistical Society (11-12 November 2013).

So a mixture of presentations, workshops and discussion.  Much like a conference, but very efficiently high level.  The subject diversity was huge and very technical too.  Here's my key generic messages from all of that.

Collaboration is key. A much stronger emerging collaboration between disciplines, and even between disciplines.  Add a call to "divide and conquer" as approaches to big problems need breaking down for the solutions, then building up the solutions.

Big data.  It's not just big it's granular and partial ("missingness").  To that I would add "now" and "open".  So big, open, now, granular and partial.  Potentially a new scientific paradigm.

Numeracy Paradox.  Recognising different reading ages and languages, maybe we need to explicitly target different numeracy levels.

Transparent Representation.  A clear desire for informed and impartial.  Prof. David Spiegelhalter's always so very engaging and eloquent here.  The example of the high profile unemployment figure which increased by 34k, with the appendix small print which gives an error rate of +/-87k, which means it could in fact have gone down.

Graphics over numbers.  Some pioneering representation of statistical risk, especially around breast cancer, which both avoids probability and an overall judgement, just presents both pro's and con's graphically - icon arrays and frequency trees.  Breaking new ground here, and can be slow going.  It's make your own mind up based on your own circumstances and  what's important to you.

Now over future.  Communicating the risk to life is evolving.  Rather than describing the risk of various heath issues as effective length of life, implicitly focussed on lost time at the end of life, the ideas now is to measure the current health age of body components.  So you might well be 40, but as a smoker your lungs have are already 50. It's about the power of now and accelerating through life, ageing faster.... that cigarette ages your lungs and extra 15 minutes right now....

Anecdotal Reasoning.  Broad aspiration to reduce that all round.

And the prize for greatest technical phrase goes to super polynomial hyperbolic relaxations, closely followed by  The Bag of little bootstraps".

A scientific paper will come in due course, and designed as a lobbying tool for more stats skills too.

Keep it Simple and Difficult

It was only 1981 when researchers asked students "How many animals of each kind did Moses take into the Ark?" to which 81% answered "two". It's only on reflection when people note that it was the arc of Noah, rather than Moses.  

This has lead to further research which focuses on this "automatic pilot" engagement, taking mental short cuts and missing key things.   That does seem to be challenging the "keep it simple"  mantra - uncomplicated, accessible and memorable - which could encourage the Moses Illusion.

Research is pointing to more effective mixed models of engagement, where short bursts of mental complexity - "cognitive dis-fluency" - can help overcome that automatic pilot.

So in a new Moses Illusion experiment, 88% went for Moses when the prose was presented in a easy to read type face, reducing to 55% when presented in a more difficult to read type face.    The more difficult type face seems to stimulate the recognition that there's a more difficult task in hand that requires, and then gets, more mental effort.  This disruption or "dis-fluency" also seems to encourage more abstract thinking.

So a multidimensional rather than uni-dimensional approach is likely to be an overall better way to engage....that balance between simplicity with some complexity.  That rather rings of the paraphrasing quote of Albert Einstein...that things should be as simple as it can be but not simpler.

Sources: Wired. Oct 2103.  Adam Alter, New York University Stern School of Business.

Statistics Definitions

Statistics is one of those terms which mean different things to different people, even within the profession.   There's good reason for that, given the wide scope from high end methodical development through to National Statistics.     In the simplest of terms a statistic can just be a data item (a datum, the singular of data), through to a number emerging from a statistical technique (such as r-squared in regression).

Source: www.statistics2013.org

Statistics becomes the process around all of that, including the analysis and interpretation.Hence there's not a specific definition in the Royal Statistical Society's new strategy.  The strap line now is data, evidence, decisions.  And "analysis" nearly had a place in there.

However for the International Year of Statistics (2013statistics.org) there's a practical set of definitions which encompass that scope:

• The science of learning from (or making sense out of) data
• The theory and methods of extracting information from observational data for solving real-world problems
• The science of uncertainty
• The quintessential interdisciplinary science
• The art of telling a story with [numerical] data

Source: A Career in Statistics: Beyond the Numbers by Gerald Hahn and Necip Doganaksoy:

So broadly the science of collecting and analysing data, and increasingly in so many areas that impact on everyday life.  Again the International Year of Stats has helpfully distil that big picture purpose too, focussing on the increasing government and business appetite for properly data driven decisions across that touch our lives in so many ways.

UK Statistical Policy Landscape Infographic

There's a strong emergence of the infographic, helping to make numbers more visual - a varied mix of shapes, colours, numbers and words.  Some are able to tell the story better than others, and variably able get over a key message or two, but none the less a very positive and welcome direction of travel.

The statistics world is going through some similarly positive moves to strengthen the impact and visibility that statistics and statistical analysis can have.   So our Royal Statistical Society has a new strategy for that, plus a stronger focus around the strap line of data-evidence-decisions.   (In fact, analysis has a part in there too and was part of the consideration).

At some point there's be some convergence between the increasingly open data that feeds lots of  the infographics, and the more long standing stats and analysis world, and especially around that Open data which is Public. 

So as a contribution to that convergence....here's my infographic which represents the current statistical policy landscape.

This was presented at the Royal Statistical Society's 2013 Annual International Conference. 2-5 September in Newcastle.

Purpose: This poster provides the macro scope, influences and connections of the UK statistical policy environment. It is also a consolidated reference source in it's on right with the key messages and emphasis emerging. It's a particular take on the ever popular infographic trend, presenting our environment in a strategic and visual way to create interest and support debate and discussion.

Abstract - This infographic presents and inter-relates the key components of the UK's statistical policy landscape.  The scope includes key bodies, legislation, policy, guidance, and key contextual factors and influences.   This provides a single strategic visual overview and reference source, illustrating the evolution, relationship and synthesis of those components over the last decade.  This also helps to illustrate to a lay audience the extent and development of the background rigour and governance in public data management, use and communication.

International Year of Statistics 2013

UK Royal Statistical Society

So 2013 is the International Year of Statistics.  Recognising the contributions of statistics to society worldwide.  It's being coordinated through statistics2013.org on behalf of the international professional statistical bodies.

It's all based around the fact that statistics have powerful and far-reaching effects on everyone, and trying to get that message over.

More than 2,000 organizations—professional statistical societies, colleges and universities, primary and secondary schools, businesses, government entities, and research institutes are participating in this worldwide event.


There's a helpful breadth of definitions of what statistics is....

• The science of learning from (or making sense out of) data

• The theory and methods of  extracting information from observational data for solving real-world problems

• The science of uncertainty

• The quintessential inter-disciplinary science

• The art of telling a story with [numerical] data

And curious and wide ranging examples of the areas in which statistics impact....

Foods you eat
Weather forecasts
Assessing disease risks
Protecting your pet’s health
Improving your health care
Transportation systems you use
Assessing your credit worthiness
Pricing your insurance policies
Ensuring national security
Examining economic health
Prosecuting criminals
Ensuring the safety of medicine
Assessing teacher effectiveness
Monitoring climate change

And predicting the need 4.4m statistical jobs in coming years, a taster for that applied work...

• Estimating the safety of nuclear power plants and alternative energy sources
• Evaluating the impact of air, water, and soil pollution
• Estimating the unemployment rate of a country
• Analysing consumer demand for products and services
• Designing studies for and analyzing data from agricultural experiments to increase crop productivity and yields

So that all has a global generic take of course, but the value shines though.  And then there's the homage to the great statisticians.....one per month over the next 12 months.

So wishing the project a fair wind and a good impact.

Source: www.statistics2013.org

Statistical Steerage

April sees the change at the helm of the UK Statistics Authority.  That can be helpful stock-take  point to get a big picture sense of where things are, and where they are going. 


UKSA was created by the Statistics and Registration Service Act of 2007 and came into being on 1st April 2008 (See Statutory Statistical).  The Authority's statutory objective is to promote and safeguard the production and publication of official statistics that serve the public good. It is also required to promote and safeguard the quality and comprehensiveness of official statistics, and ensure good practice in relation to official statistics. 

Interestly, the Chair of UKSA is Crown appointment, made with the approval of Parliament following a pre-appointment hearing before the House of Commons Public Administration Select Committee and a formal motion debated on the floor of the House of Commons.  The first five years have been under the stewardship of Sir Michael Scholar KCB.  The baton has now been handed over to Andrew Dilnot CBE.  An eminent public figure (Institute of Fiscal Studies, Social Security Advisory Committee, the National Consumer Council)  and indeed the founding presenter of  Radio 4's more or less show about stats in real life issues.  A good provenance then. 

So where are we.  In short the UKSA has been doing three things:

1. Code of Practice for Official Statistics.  Creating what is in essence the guide to best practice context for data - consolidating, refining and developing lots of guidance over the decade from amongst others, the Audit Commission, HM Treasury, Cabinet Office.  Covering 74 business processes (principles, protocols, practices) from consultation on needs through to publication processes.  It is quite a challenge to get right, and helpfully sets the bar high.  In the spirit of the original Chinese saying...."If you aim for the top of the tree you'll probably never leave the ground, but if you aim for the moon you might at least get to the top of the tree".  Fundamental here is that this is a generic code, and applicable to "official" as opposed to just "national" statistics.  Some interesting emphasis has follow on this.  


2. National Statistics Stamp.  Steadily working through the entire scope of national statistics and given them the stamp of approval, resulting in:


- Assessment Reports: Currently 190 and counting. 
- Breach Reports in respect of the Code: 59.  
- Issues Log for issues highlighted to UKSA in 11/12: 50


3. Challenging and championing.  In respect of the use of official statistics.  Leading and supporting some of the very technical debates (Retail Prices Index etc.) and at the other end of the spectrum. very publically challenging the inappropriate use of statistics by those who should know better.  Ministerial apologies being a familiar outcome.


So where are we going.  Perhaps most indicative at such a pivot point are the parting shots from the previous chair and the very early signs from the new chair.


A. Administrative data is an opportunity


This is a significant and substantive outgoing message.  It's all about the use of administrative date for statistical purposes.  Building on some earlier debate about the boundaries of management information, statistical, research administrative data. In short how administrative data - process data collected not primarily for statistical purposes - can be effectively embraced and harnessed for statistical use, with the right expert engagement and with important caveats too.
   
The wordcloud of the UKSA letter to the Cabinet Office on the use of administrative data....

In fact data is data.  It's all of the other things which are different and determinate. Purpose, definition, methodology, accuracy, scope and so on  (See Data Provenance).  Understanding that context is the key to unlocking whatever the potential might be.  So having compartmentalised data to help clarify and understand, there is a timely opportunity to bind them together in an informed and useful way.  


B. Interpretation is key.


The UKSA has created a foundation to explore the importance of the role of interpretation. After all the data the means to an end - specifically the means to the message.  This has been encouraged and supported by the influential Public Administration Select Committee "recommends that the Statistics Authority should take a proactive role in ensuring that data released is intelligible, objectively interpreted and in a readily accessible format"


This is also emphasised in some of the correspondence to MPs encouraging "more systematic public presentation of all the relevant statistical material along with appropriate professional explanation and guidance".   This was reflected more generally in the significant and succinct UKSA guidance on The Value of Statistical Commentary, previously discussed in Age of Analysis.


This is clearly echoed by the incoming chair, in the second sentence of the initial statement... "I care.... not just about the numbers themselves but about the ways in which they are used and interpreted", and "the need and value of sound and strong statistical analysis" and that "impartial commentary, explanation, and tools for dissemination are critical to our success".   Here's the word-cloud for that opening statement...

So the debate starts to move beyond data into the surrounding analytical and interpretation.  That said the early emerging focus is around what might be described "interpretation of tables"
rather than "process of analysis" from which insight and messages are distilled and synthesised, and of course where the real value is realised. A means to a greater end (see Impactometer).

Hence data is the foundation, and the debate progresses to value the analytical processes that build on and around that.  Probably not helped by the ease by which the term statistics is interpreted so differently  - from "the quantitative data" through to the "science concerned with the collection, classification and interpretation of quantitative data".  So statistics is both the data and the interpretation.  Of course, not forgetting the specific statistical techniques which can produce statistics.....