I agree with twotribes and fatwolfie.
Getting unbiased data and end results takes care. It is easy to introduce bias either deliberately or accidentally.
Whoever produces the stats is important too as they can conduct trials with built in bias.
Collecting unbiased data is step 1.
Step 2 is getting correlation to show links/cause and effect is also open to errors. There can be too many influences that need to be eliminated or they must feature equally in the comparisons made.
It is worries about lack of correct data and poor comparison that makes me suspicious, especially when politicians are involved. Diesel was good at one time and promoted by politicians for better mpg and CO2 , ignoring NOx at that time. Now the NOx has reared its head reversing earlier decisions.
