I have access to a database where people declare how far they walked on a reference day. Walking inside buildings or within facilities such as malls or campuses is not included in the estimate. Calculated in this way, the average distance walked per person is 1.7 km per day. I am creating categories and sub-categories based on the distance walked: 0-1 km, 1-2 km, etc. The question is if and where I should cap people's responses. Some seem unrealistic, around 50-70 km on the reference day, but I need to define a cut-off point based on something scientific. The setting is Switzerland (whole country: urban, surburban and rural) and the data are in kilometres. Of course, mile-based calculations or contributions are also welcome. Many thanks in advance for your help.
Spinney et al (Journal of Physical Activity and Health, 2012, 9, 153-162) provide some useful data based on time diaries in Halifax, Canada. It's important to note that on any given day only 32% of respondents took an active-transport walk, and only 22% took a recreational walk. For those taking each type of walk, the mean times for daily walking were 23 minutes of AT walking and 51 minutes of recreational walking. However, the medians were only 15 and 44, whereas the upper quartiles were 31 and 64. So, there is a skewed distribution.
In a separate study, Millward and Spinney (Journal of Transport Geography 19 (2011) 51–58) report a mean daily duration of all walks for all respondents of 15 mins/day, and for those reporting walking (i.e., participants) of 37 mins. This suggests a combined upper quartile (for all walking) of around 48 mins/day. Thus, anyone walking over, say, 75 mins/day would be an outlier.
For translating time durations into distances, Millward et al. (Journal of Transport Geography 28 (2013) 101–110, table 1) report mean speeds for AT walks of 4.8 Km/h, and for recreational walks of 4.3 km/h.
Dear Hugh, many thanks for your answer. I am a great fan of these papers and now I am going to read them (again!) :-)
Would your estimate imply that "outliers" begin at 1.5 times the lower limit of the upper quartile? Or maybe I misunderstood.
Yes, Derek, 75 minutes is approx. 1.5 times the upper quartile for participants (those who walk on a given day), although it's perhaps 3 times the upper quartile for all respondents. I haven't run the numbers, but I would suggest that 75 mins. besides being a nice round number, is also approx. the 95th percentile.
- Badges
- Science topic
- Similar topics
- Statistics
- Survey