Add comment on reliability

vignettes/report2.Rmd

@@ -195,6 +195,11 @@ knitr::include_graphics("https://raw.githubusercontent.com/saferactive/saferacti
 
 <!-- The results of this GAM model can be applied to Local Authorities.... -->
 
+## Reliability of count data
+
+The analysis above shows that the although the count data is 'noisy', it has potential to provide an indicator of *relative* change in cycling levels at regional to local authority levels.
+A next step, outlined in the final section of this report, is to assess how large the confidence intervals are associated with this dataset, by comparing the relatively sparse DfT count data with larger external datasets including TfL's open cycle counter network data.
+
 # Geographic data analysis
 
 Building on methods we initially developed for the [CyIPT](https://www.cyipt.bike/) project, we developed new techniques for allocating crashes to features on the road network, such as junctions and road sections.