Road Safety News

Road Safety GB announces date and venue for new data analystsí national conference

Wednesday 2nd December 2015

Road Safety GB has unveiled plans to stage a national conference for road safety data analysts in London in March 2016.

The conference, which will be organised by Road Safety GB in partnership with Road Safety Analysis, and with support from the DfT, will become an annual event.

The conference is a result of a survey conducted earlier this year which captured the views, experiences and aspirations of more than 100 analysts from across the country.

The survey findings, which were published in a report, showed that while respondents are confident in their analytical skills, there is a clear desire among analysts for further training and for sharing best practice.

Among the report’s recommendations was a national conference specifically focussed on supporting those in the road safety profession who carry out some analytical work as part of their role.

The inaugural conference will take place at the Royal College of Surgeons, Lincoln’s Inn Field in central London on 1 March 2016.

There will also be the opportunity for a small number of organisations to exhibit alongside the conference.

Honor Byford, chair of Road Safety GB, said: “Our management team has studied the recommendations in the report and we are delighted to be able to take some concrete steps to support colleagues around the country with an event that will engage and equip professionals for their work.

"This event will be designed to bring together collision and casualty data analysts and those who work in connected fields - including fire & rescue services, academic institutions and public health - to network and share knowledge, good practice and research relating to their specialist role."

More details about the conference will be published on the Road Safety GB website in the coming weeks and the booking process will open shortly.

In the meantime, click here to register interest in attending either as a delegate or exhibitor, or for more information contact Dan Campsall at Road Safety Analysis on 01295 731812.


Comment on this story
Report a reader comment

What's your view - comment on this story:

I confirm that I have read and accept the moderation policy and house rules relating to comments posted on this website.
Your comment:
Your name and location:
Your email:

Hugh and Idris
I suspect this discussion between the two of you may run on, so I'm going to intervene and thank you both for your contributions but draw a line under it at this point.
Nick Rawlings, editor, Road Safety News

Agree (7) | Disagree (2)

The number of fatalities do not conveniently match or track the number of collisions, so increases or reductions in fatalities do not reflect corresponding increases or reductions in collisions which, in turn directly resulted (or not) from behavioural changes, in turn influenced by interventions.

If one is trying to determine the success of a collision prevention intervention, you have to judge it by the number of incidents you've hopefully reduced that would otherwise have happened and not by the consequences of the collisions, which can be due to many unrelated factors. The number of collisions prevented by enforcement (including speed) is actually incalcuable.
Hugh Jones, Cheshire

Agree (1) | Disagree (4)

First, a correction -"...that 27% covers only those accidents that those involved CANNOT avoid becoming known to the authorities"

Hugh - you seem not to understand my point - fatality numbers are recorded very accurately so when SI consistently tracks fatalities closely, that trend is likely to be accurate too. And as engineers know, it is the RATE OF CHANGE of a parameter - in this case KSI - that measures success or failure. So your the basis of your "meaningless" claim remains wrong.

The DfT's 27% estimated reporting levels gives us reasonable figures for the real totals.
Idris Francis Fight Back With Facts Petersfield

Agree (5) | Disagree (0)

Idris: If one is only concerned with measuring deaths, then the available data may suffice, but I thought we (i.e the road safety profession) was concerned with preventing collisons in the first place, in which case, not knowing the actual number of collisions which occur and which we're trying to prevent, does make the exercise meaningless I'm afraid.

Under-reporting is not necessarily due to people not wanting to involve the police - I've been at collision scenes where the police attended, but as there were no casualties (as in my example earlier) they did not bother with any official report or investigation.

It's like medical scientists claiming they must have found a cure for the common cold because there is no official record of anyone having one.
Hugh Jones, Cheshire

Agree (0) | Disagree (8)

Hugh - we have no alternative but to work with the best data that is available, while acknowledging that, as you say, reporting levels of non-fatal accidents are low.

However reporting of fatal accidents is, for obvious reasons, close to 100% and those who have taken the trouble to obtain the data know that reported SI tracked K remarkably accurately at a ratio of about 10:1 from 1962 to 1998 - see for example It follows that although SI numbers were certainly understated throughout that period the trend (which is what matters) WAS accurate.

The same graph shows a very unusual sustained deviation of K from its long-term trend (it all-but stopped falling) due to the simultaneous and sustained "longest boom we have ever seen" from 1993 to the crash of 2007.

While it was logically inevitable that SI reductions slowed to a crawl at the same time, that the graph shows them continuing to fall was due largely or entirely to a reduction in SI reporting, as the DfT confirmed, from an estimated 1 in 2.7 to 1 in 3.7 over that period.

That was probably the last significant fall in SI reporting levels because that 27% covers only those accidents that those involved can avoid becoming known to the authorities.

For all these reasons Hugh, the premise of your argument that any analysis is "meaningless" is simply wrong. That so many drivers seem to prefer to avoid involving the police if they possibly can has serious implications for society in these troubled times.
Idris Francis Fight Back With Facts Petersfield

Agree (6) | Disagree (0)

Triggered by Idris's comment (and Dave Finney's) and following my earlier comments, perhaps the conference's priority should be to discuss the usefulness of comparing and analysing collision 'data' which, due to significant under-reporting is inaccurate in the first place. The data does not give the ACTUAL number of collisions anywhere, only the reported ones.

Doing before and after comparisons at any sort of accident site is therefore inconclusive and regression to the mean and other statistical factors become meaningless, unless you're working with numbers which are known to be true in the first place.
Hugh Jones, Cheshire

Agree (3) | Disagree (6)

Like Dave I welcome this Conference and wish to give a presentation on speed camera analysis. I am however puzzled that Dave wishes to quantify as well as to exclude the effects of regression to mean (RTM).

His 2010/14 analysis of speed cameras established from the data what all previous analysts missed (though centuries' old laws of chance make inevitable*) that the entire RTM fall occurs the moment biased (i.e. minimum KSI rate) site selection periods end. The inevitable delay before cameras can be installed and have any effect therefore allows us to differentiate between the two effects on the basis of timing alone.

In any case, the scale of the RTM fall, being an automatic and exact correction of prior selection bias, is a function of the volatility and in turn the scale of the raw data, and also of the degree of selection bias. As all these factors vary widely from site to site and from any one group of sites to any other, so will the RTM % fall, and it would therefore futile to try to determine its value.

* For example, 6000 people throw dice. The laws of chance ensure that very close to 1,000 will score 1, another 1,000 score 2 etc. Select the 1,000 who scored 6 to throw again. Do they all score 6 again? Of course not, close to 166 score 1, 166 score 2 etc. In other words, abnormal scores selected for being abnormal at a particular time instantly return to normal the moment that biased selection ends. The same applies to camera sites - as my own analysis of monthly data from thousands of cameras sites confirms - and in many other contexts.
Idris Francis Fight Back With Facts Petersfield

Agree (5) | Disagree (4)

Following my last comment, on those occasions where the police did attend a collison and write up a report, wouldn't it be interesting to compare their (official) version with the claimant's version?...not least to see if the claimant's excuse/mitigation, matched the police's idea of the contributing factors.
Hugh Jones, Cheshire

Agree (7) | Disagree (0)

I'm with Hugh in imagining what would be revealed if we could pick through all those insurance claim forms. To have such a huge resource in place and yet not be able to get at it shows that there's something terribly wrong somewhere.
Duncan MacKillop. No surprise - No accident.

Agree (8) | Disagree (0)

How about inviting representatives from the motor insurance industry along with a view to them being encouraged to provide information on those collisions (the majority reputedly), that currently do not appear on the official collision records that are being analysed? The causation factors are the same - only the consequences differ and it is these that determine whether they appear on Stats 19 and give a distorted picture of what's happening on the roads.

Recently there was a horrendous looking collision at a junction locally, where at least one vehicle was damaged almost beyond recognition but - no injuries, so no official record of it ever happening. That could theoretically happen several times and yet the junction would show up as having a 'good accident record'!
Hugh Jones, Cheshire

Agree (8) | Disagree (0)

This conference is an excellent new idea from Road Safety GB and the timing could not be better. There are a host of interventions that desperately need good quality analysis (such as 20mph speed limits, vehicle-tracking devices for insurance and eCall, graduated licences etc) and there is also the new requirement that authorities publish "the impact of speed cameras on road safety". Separating the effects of these interventions from all the other factors will be a major challenge.

The effect of "selection" (often known as RTM) is potentially the largest effect that confounds results and is my main area of interest. I would be willing to put together a presentation on what RTM is, why it occurs, and how to measure and exclude it's effect, if required.

I wish Road Safety GB all the best and I am sure attendees will find the conference to be of great benefit.
Dave Finney, Slough

Agree (10) | Disagree (5)