0

I recall various conversations with biologists, ecologists, and foresters that I neglected to ask for clarification on at the time. It doesn't occur in any of my statistics references.

Sometimes in reference to methodology I've heard people use the phrase "skewing the data" or "skewing the results". The specifics have pretty much gone from my memory, but here are some examples of the form of what was said.

  • We avoided using <reagent_X. on the samples to prevent skewing the results.
  • We sampled sites with only <property_Y> to prevent skewing the data.

While I wouldn't say a priori that they couldn't be talking about skewness in some statistical sense, my intution is that they were not. Different disciplines are allowed to invent their own terms for things, but sometimes I think we're talking about the same sorts of things using different words. That is not really such a sin either, but it is (in my opinion) convenient when the terminology converges when it is discovered that two notions are synonymous.

Example of Converging Terminology

An example of this being somewhat successful is with canonical polyadic decomposition which was originally devised mathematically as "canonical polyadic decomposition" but was also later reinvented as "PARAFAC" and "CANDECOMP". Many people now refer to it as just "tensor decomposition" in some contexts, although "canonical polyadic decomposition" is still used when we wish to distinguish it from other functional decompositions of tensors such as Tucker decomposition or tensor train decomposition.

My interpretation is that my interlocutors were referring to what I would call "selection bias", but it seems to be rather difficult to find direct explanations of the phrases "skewing the data" and "skewing the results".

Is "skewing the data" and "skewing the results" just selection bias?

Galen
  • 8,442
  • 1
    Please see remarks at https://stats.stackexchange.com/questions/202879/what-misused-statistical-terms-are-worth-correcting/202886#202886 In short, skewing here can only mean biasing for your examples to make sense. – Nick Cox Aug 27 '23 at 15:55
  • 1
    In my experience scientists and engineers use "skewing" in various ways that often lack any formal definition or clarity. They might mean focusing samples on subjects with particular properties; or obtaining relatively more samples for such subjects; or using biased procedures; or many other things. – whuber Aug 27 '23 at 17:20
  • 1
    From your quotes, they seem to mean either to avoid introducing some form of bias or to focus on some specific comparison or estimate rather than contaminate the comparison of interest with some other subset(s) of a broader population (again, avoiding some form of bias). However it's difficult to infer the content of people's heads from brief, likely paraphrased second-hand reports of what they said from the past, without the specific context. We certainly have less opportunity to seek clarification of what they meant than you had. – Glen_b Aug 28 '23 at 00:06

0 Answers0