“Leo Breiman was still a statistician, and while he critiques parts of his profession he’s not dismissing the importance of statistics. The same can be said for David Donoho’s excellent recent article, “50 Years of Data Science.” The fact that some statisticians have too narrow a scope in no way diminishes the validity and importance of statistics to good decisions and valid science. Sure, it is incumbent upon statisticians to prove their worth to the larger, emerging discussions around data science, but I remain concerned about “data scientists” who may be able to wrangle data well and write code but don’t understand the fundamentals of analysis.” Polly Mitchell-Guthrie
Myth #13: Statisticians are homogeneous
Myth #14: Academic statisticians are typical of and can speak for the whole profession
First, statisticians are not homogeneous.
Instead, there is a rich diversity of applied statisticians/quants and they work on every conceivable data analysis problem and in every technically advancing field. A statistics degree is not required; knowledge of applied statistics and the domain are essential.
Sometimes, we perceive groups that are unfamiliar to us as homogeneous. Table 1 explores pockets of statisticians in no particular order. The descriptions were not robustly collected; there are no comprehensive information sources, no surveys of what is going on with applied statistics in the field.
|
Type |
Description |
Organizations |
|
Clinical Statisticians/ Biostat |
Masters of Pharmacology, Clinical DoE, and FDA regulations; publishes results of clinical trials |
ENAR, WNAR, et al. |
|
Government Statisticians |
Masters of Sampling; Portmanteau Statisticians |
ASA |
|
Fellowed Statisticians |
Possesses ‘fellow’ like appointments in think tanks at not-for-profits or large corporations |
Depends on industry |
|
Industrial Statisticians |
Masters of QC & DoE; works with engineers |
ASQ |
|
Academic Statisticians |
Publish or perish; many at research universities; emphasizes coefficient estimation & parametric statistics; majority of statistics professors |
ASA |
|
Teaching Statisticians |
Dedicated to teaching first; minority of professors |
ASA |
|
Operations Researchers |
Works on OR problemsmany industries: shipping, military, government, et al. |
INFORMS |
|
Financial Statisticians/ Econometricians |
Econometricians; may require mastery of finance, regulations, insurance, accounting, et al. |
Banking |
|
Corporate Statisticians/ Marketing Science |
Some Portmanteau Statisticians; Masters of Prediction; requires business savvy, communication, leadership, and people skills |
Depends on industry |
|
Niche Statisticians |
Applied statisticians in numerous fields like Energy, Insurance, Mining, et al. |
Depends on industry |
|
Computational Statisticians |
Software development and similar |
Software industry, et al. |
Table 1: Pockets Of Statisticians
Many applied statisticians identify with their application domain first and statistics second. They often have a degree in another field and are engaged with industry specific journals, conferences, and organizations. No matter how strictly, yet reasonably, we define ‘applied statisticians,’ there are far more of them than the 18,000 members of ASA.
Second, academic statisticians are not typical of and can not speak for the whole profession. The most self-absorbed academic statisticians have made the twin mistakes of trying to speak for the whole profession and not realizing the extent and distinction of applied statistics. This is analogous to talking-head MDs, who are paid to discuss the best ‘standard of care’ for a new pharmaceutical product, when they lack both a medical practice with real patients and the humility to not lecture outside their practical experience.
At their best, academic statisticians are tool builders and applied statisticians are problem solvers. Both ascend from the same set of early statisticians, who analyzed data and wrote about it. There are growing differences between the two groups.
As academic fields mature, e.g., economics, mathematics, physics, et al., they become less involved with applied problems and methodology, and more focused on publishing theory. E.g., in ‘Boom, Bust, Boom’ (2015), they make the point that academic economics is overly focused on neoclassical economics and does not adequately address real-world market crashes. When economics students graduate, they lack the tools to explain or model a market crash and, like graduating statistics students, still have many years of self-training ahead of them.
Table 2 contrasts these two groups of statisticians. The most academic of academic statistics is to publish a new theoretical idea and this is likely to involve proofs rather than data. The most applied of applied statistics is to derive a new solution to a problem and this will involve data and no new proofs. Decades ago, there was much more overlap between these two endeavors.
|
Academic Statisticians |
Applied Statisticians |
|
Publish or perish |
Analyze or perish |
|
Killer App: New tools |
Killer App: New applications |
|
Ascended from early statisticians, who wrote about analyzing data |
Ascended from early statisticians, who solved problems using data analysis |
|
Publications might not contain data |
Data analysis unlikely to be published |
|
Might not have analyzed data before |
Might not have published before |
|
Reticent to ‘go native’ (learn other domains) |
Must master other domains |
|
It’s about the math and logic |
It’s about the people and the domain |
|
Builds tools |
Adapts tools |
|
Focused on parametric statistics |
Focused on problems within their domain |
|
Wants to sell as many degrees as possible (to support their publication habit) |
Stop selling so many degrees (this cheapens the profession) |
|
Favorite Hobby: Teach everyone parametric tools and theory |
Professional Norm: Keep techniques proprietary; share with other quants/statisticians sparingly |
|
Practices with synthetic laboratory data, which is relatively agnostic to the domain |
Faces real messed-up data within domain context & with domain assumptions |
Table 2: Academic Versus Applied Statisticians
The academic literature is an encyclopedia of promising ideas, some have been impactful, others … remain to be seen. Success in publishing comes with it the myopia of specializing in particular tools. The corresponding academic incentive structure over emphasizes publishing at the expense of everything else. E.g., it has been said that the best researchers make the worst teachers.
One tactic of anti-statisticians is to leverage statistics professors to mischaracterize applied statistics. They either misquote them or misattribute their comments about academic statistics to applied statistics. When most academic statisticians talk about statistics they are talking about academic statistics only.
Close
Statisticians are heterogeneous; some of the differences stem from the breadth of domains within which they reside. Even with the best intentions, academic statisticians can not represent or speak for applied statisticians.
Conclusion of series: Unless a summary blog is wanted, this will conclude the 13-blog series on Statistical Denial. No matter the means: renaming, mischaracterizing, or fabricating, the result of statistical denial can be to exclude the proper qualifications, diagnostics, review, thinking, assumptions, and tools.
We sure could use Deming, right now. Many of us who embrace the explicit rigorous logic and protocols of these tenets of data analysis hang out in the new LinkedIn group, About Data Analysis. Come see us.
The entire Statistical Denial series can be found on Datafloq.
- Statistical Denial 1, Blog 1: Essays On Statistics Denial
- Statistical Denial 2: Statistics Debacles & The Coming Flood Of Statistical Malfeasance
- Statistical Denial 3: Applied Statistics Is A Way Of Thinking, Not Just A Toolbox
- Statistical Denial 4: Five Forces Pushing Statistics Expertise Out of Data Analysis
- Statistical Denial 5, Myth 1: Traditional Techniques Straw Man
- Statistical Denial 6, Myth 2: Why Statisticians Not Only Practice Within Traditional Statistics
- Statistical Denial 7, Myth 3: Are Data Mining and Machine Learning Distinct from Statistics?
- Statistical Denial 8, Myth 4: Why Prediction Is / Is Not Part of Statistics
- Statistical Denial 9, Myth 5 and 6: Why Statistical Significance Does Work For Big Data
- Statistical Denial 10, Myths 7, 8 and 9: 3 Statistics Denial Myths on the Volume of Big Data
- Statistical Denial 11, Myths 10 and 11: Debunking the Myth that None of Statistics Works for Big Data
- Statistical Denial 12, Myth 12: Publications Straw Man
- Statistical Denial 13, Myths 13 and 14: Minimizing The Profession
