- Joined
- Mar 5, 2018
- Messages
- 8,009
- Reaction score
- 1,428
- Location
- Seattle, WA
- Gender
- Male
- Political Leaning
- Very Conservative
Changes such as those you mention are exactly the reason why old temperature readings are adjusted. A large part of the reconstruction of historical temperature records involves the development and use of various methods for correcting for such changes.
You cannot use cooked data in statistical analysis. It is useless. No analysis has been run yet. There is nothing to correct by. You cannot use the output of an analysis as it's own input! You can't even use it for the next time you perform an analysis.
Raw data MUST be used. Selection MUST occur by randN. That means if a point is selected, it must be ONCE and only ONCE. Not twice, not 0.7 times, ONCE.