I came across the term "uncomfortable science" on Wikipedia (https://en.wikipedia.org/wiki/Uncomfortable_science) where it is defined:
- Situations where the same data is used for model development (exploratory data analysis) and model testing (confirmatory data analysis).
I found the book where the term was first defined (Tukey, J. 1954: Unsolved Problems of Experimental Statistics, Journal of the American Statistical Association, Vol. 49, No. 268, bottom page 718). Here it says:
Statisticians must face up to the existence and varying importance of systematic errors. The failure of the statistician to take sufficient cognizance of systematic errors has been in part an escape phenomenon. To a man looking hopefully for a way to shorten a confidence interval by 7 per cent of its length by ingenious devices, the thought of systematic errors which might make it twice as long comes as a severe shock, and all men try to avoid shocks. Perhaps, too, the recent development of statistics in connection with the uncomfortable sciences like agriculture and biology -uncomfortable because unsystematic errors tend to be so large - may have much to do with this. Only the sampling survey statisticians, with their recent treatment of "non-sampling errors" seem to be facing up to the existence of systematic errors.
From this I understand the definition of uncomfortable science being:
- Situations where random error is so large that systematic errors cannot be identified.
However, I have only seen examples of texts using the former definition.
My question is: Are these two definitions the same? And if so, how are they linked?