The first hypothesis here is that crashes on the weekend are likely to be more fatal. The previous bar-chart had shown that Sunday had the lowest number of fatal crashes by day of week, yet the above tells us that they are more likely to be fatal – because there will be correspondingly far fewer slight crashes recorded. Whilst this australia contact data may be a correct insight, an alternative data-collection based explanation could be, ‘there are less police officers on duty over a weekend and less slight crashes are therefore recorded than should be’, which would also lead to this result. We therefore need to understand the limitations of the data collection as well as the context of the data in knowing when we can draw conclusions from our analysis.
Similarly, if we banded the above by the actual day of the year, we would find a pattern that the days around Christmas seemed to be the time of the year when we would expect most fatal crashes. This could be true, but this also seems to suggest that it could be that fewer than expected slight crashes are recorded in periods when more police officers would be on holiday.
These results should make us stop and think quite carefully about drawing any insights about factors that are important in determining whether a crash is likely to be fatal.