Issue Date: January 29, 2018
Letters to the editor
I read with interest your recent guest editorial by Richard Harris on reproducibility issues in science and their impact on research findings (C&EN, Nov. 27, 2017, page 2). I agree with the premise that statistical methods are abused, misused, and often poorly understood by experimentalists, including chemists.
My hypothesis to explain this predicament is that most, if not nearly all, undergraduate chemistry departments do not require chemistry majors to take statistics classes during the course of their studies. Statistics may be listed among the recommended electives but is otherwise largely ignored. The problem of statistical ignorance is further exacerbated in graduate school, where again no such requirement exists. As a consequence, chemists such as myself who wish to employ these methods are forced to “learn” them on the job, thereby perpetuating statistical blasphemy.
In following the time-honored tradition of statistical abuse, I examined the undergraduate degree requirements of three highly regarded institutions to test my hypothesis. Because the data collected from my brief investigation agreed with my hypothesis, I therefore conclude with unequivocal confidence that my hypothesis is correct! Never mind the fact that my data set was much too small and not representative of the population; the scientific literature has taught me that these issues are statistically insignificant when compared to unprecedented, robust, and novel findings.
Bryan R. Moser
The reproducibility of synthetic organic procedures has always been an issue. “I didn’t get the same yield; the same purity; the same product—as before.” The simple reason is that synthetic chemists may not know exactly what they did during a given procedure.
“Refluxed for five hours” tells nothing about the rate of heating. “Added over two hours” does not mention that the stopcock was continually adjusted. “Cooled in an ice bath” leads to a different cooling rate for a larger reaction. Lack of data-logging devices means that exotherms due to accumulation of unreacted materials are easily missed when the chemist steps out for a cup of coffee. The solution to these issues is to log reaction data and use automated reaction control.
I can cite specific examples where I tried to reproduce a literature reaction while data logging and discovered temperature excursions—once to the point of personal danger—that the originating chemists never documented. Maybe they stepped out for a cup of coffee.
Prior to my retirement from a kilogram scale-up lab, I demonstrated that, with the same batches of starting material, the yield and purity of product was guaranteed to be reproducible from the gram to kilogram scale within the limits of measurement error. How did we do this?
First, we documented small-scale reactions via data loggers. We then ran our reactions on in-house automated reactor systems as well as the Contraves automated reactor, the Mettler LabMax, and the Mettler RC1.
The biggest impediments to wide adoption of data logging and automation were crusty old chemists who, and I quote, said things like: “I can do anything I need to do in a beaker on a hot plate.”
A reader responded to a Newscripts article on oysters’ hearing abilities (Dec. 18, 2017, page 56) with a rhyme.
My late father was fond of wordplay. Here is one of his I remember:
“What kind of a noise annoys an oyster?
A noisy noise annoys an oyster.”
Jan. 8, page 6: In a news article about sulfones being used in cross-couplings, commenter Scott Denmark mentions the unique reactivity of N-phenyltriazolylsulfones. The reagents are actually N-phenyltetrazole sulfones. In addition, because of a production error, the reaction scheme in the story was cut off, showing only part of this reagent. Here is the correct structure.
Jan. 22, page 22: A feature about the start-up company Numaferm contained an incorrect description of peptides. Peptides are amino acid chains, not DNA chains.
- Chemical & Engineering News
- ISSN 0009-2347
- Copyright © American Chemical Society