Comments on report 3

[joeytalbot]

No description provided.

[joeytalbot]

Please change the colour scheme on figure 1.1 I can't distinguish all those blues.

[Robinlovelace]

Tricky to do as it's a screenshot! https://www.ons.gov.uk/peoplepopulationandcommunity/healthandsocialcare/conditionsanddiseases/bulletins/coronavirustheukeconomyandsocietyfasterindicators/10december2020

[joeytalbot]

Line 151: what does 'and where selected for special study' mean?

[Robinlovelace]

Good question, clearly a typo. Well spotted!

[joeytalbot]

Figure 2.2 - it would be useful to have a close up of London, since that is our study area and it is not visible.

Also, it looks like some areas on this map have high casualty rates because of a high tourist/day tripper population i.e. people who are neither residents nor workers. This could include Cumbria, the Scottish Highlands, North Wales and the South coast.

[Robinlovelace]

Line 151: what does 'and where selected for special study' mean?

Heads-up, I've proposed a fix to that, see here:
#61

Does that look good to you @mem48 (who originally wrote that ; ) and @joeytalbot who spotted the issue.

[Robinlovelace]

Figure 2.2 - it would be useful to have a close up of London, since that is our study area and it is not visible.

Good point but I think this section is more about the big picture. Would it be easy to add an inset map of London @mem48 ? Maybe worth doing if so.

[Robinlovelace]

Also, it looks like some areas on this map have high casualty rates because of a high tourist/day tripper population i.e. people who are neither residents nor workers. This could include Cumbria, the Scottish Highlands, North Wales and the South coast.

Agreed, that is commented on in the text. Any further changes to highlight that are welcome though.

[Robinlovelace]

Is the deadline Tomorrow or the meeting is tomorrow?

Meeting is tomorrow. Comments and changes to the report welcome whenever. Many thanks!

[layik]

Sorry @Robinlovelace just deleted that. Great, I will add something to the report, please discard if not ready/immature.

[joeytalbot]

Line 239: how could high crash rates per km be an artefact of low cycling levels. High variability in crash rates could be artefact, but high crash rates suggests a real problem.

The more likely artefact is in differing proportions of commuter/leisure/personal cycling activity. i.e. models of cycling activity based on travel to work don't capture so much of the cycling in areas where non-commuter cycling dominates.

[joeytalbot]

Figure 2.7: it would be really nice to have the London Borough visible so we can see what's going on better.

[Robinlovelace]

Line 239: how could high crash rates per km be an artefact of low cycling levels. High variability in crash rates could be artefact, but high crash rates suggests a real problem.

Good point. I meant that if you have one crash in an area with very low 'exposure' (cycling in this case) that area could be seen as risky when in fact it was just a chance event. I agree with your point that not all crashes during peak hours are commuting though. How would you rephrase it?

[joeytalbot]

It is clear from the results that outer London is a more dangerous place to cycle than inner London.
However, places where cycling levels are low will tend to have more variable collision rates with less certainty, which means that an apparent high rate in a given year could actually be an artefact of the high variability in collision rates per km cycled.
There are at least three ways of dealing with this issue:

• Only focus on areas with a sufficiently high level of exposure (cycling km in this case), with some cut-off (e.g. 1000 km per 500m grid cell per year minimum).
• Geographic aggregation, as shown in the figures above.
• Visualisation or other techniques to communicate uncertainty, e.g. by making estimates based on lower levels of exposure more transparent.

A second potential source of error is that, even in peak hours, many cycle journeys will be for non-commute purposes.
The proportion of journeys that are for the purpose of travel to work will vary from one area to another.
Since we are only counting journeys to work, our estimated collision rates per km cycled will be relatively higher than they should be in areas where less cycle journeys are for commuting purposes.

[Robinlovelace]

Is the text in your last comment a suggested change @joeytalbot ? Please put in a PR by editing here if you get a chance:

https://github.com/saferactive/trafficalmr/edit/master/vignettes/report3.Rmd

Otherwise I'm happy to make the change. Many thanks 👍