Most of the studies on climatic trends adopt a confidence level of 90 or 95%, but don't mention any reason for the choice. What do you think should be the strategy to choose this value? Can we say that 95% is too conservative a choice?
For most common hypothesis:
H0: trend is zero (none) versus H1: exist any trend
the 95% confidence intervals are approprite (if hypothesis is tested on standard alpha level of 5%, 90% CI are related with alpha level of 10%)
But if equivalence hypothesis is tested
H0: absolute value of trend (absolute value because in general trend could be up or down) is above some threshold (e.g. 1%)
versus
H1: trend is withing so called equivalence range < - X%, + X% > (I am not working in climatology area therefore I know which X% are acceptable for different medical questions ... the percentage X% is not statistical issue, this threshold has to be selected by experts from studied area - here experts from climatology what is somehow equivalent)
then 90% CI has to be completely within < - X%, + X% > if hypothesis is tested on alpha level of 5%.
So I think using of 90 or 95%, but not mentioning any reason for the choice ... it should be related with hypothesis selection as I explained above.
Thank you Ladislav Pecen for the comprehensive response. What do you think about the last part of my question? Do you think 95% CI is too conservative a choice when dealing with natural systems like climate or plants? Cheers.
- Badges
- Science topic
- Similar topics
- Mathematics
- Statistics