Sun, Apr 22, 2018 - Page 6 News List


Bad academic statistics

The article on academic papers makes me wish that fewer, but clearer analyses were published occasionally (“Fewer, but better papers published: survey,” April 14, page 2).

The information presented compares the number of papers published by Taiwanese academics between 2012 and 2016 (138,411, according to the authors of the report) with the number published by the same group between the years 2011 and 2015 (139,408, from the same source). It should be clear that the years 2012, 2013, 2014 and 2015 are included in both groups.

Anybody with a minimal understanding of statistics should know that the totals are not independent. If they were independent, the difference between them would not be “statistically significant,” meaning that differences in two non-overlapping periods of five years could differ by more than the observed difference, about two times out of 100.

Obviously, the authors of the published analysis had available the number of articles by year (or could have made arrangements at minimal cost), and something more intelligent might have been said from analyzing the trend in those numbers.

What is worse is that the whole discussion pays no attention to the steps that lead to those numbers. One might think of the process as very complex, but we might simplify it at least as follows:

One, authors submit papers to journals.

Two, each journal decides what articles to publish, a process that involves the standards of the journal, as well as the relation between the number of articles submitted and the number of pages the journal needs to fill.

When we consider that simplified process, some considerations that have no place in the oversimplified “things just get published” version come to mind.

The number of papers published by Taiwanese authors might have declined because the authors submitted them to journals of higher standards, but, equally well, they might have declined because more authors from other countries submitted more papers to the journals.

I would not expect the authors of the evaluation to have access to the data required for a conclusive analysis, as journals do not publicize the number of articles submitted by researchers of various countries, but the authors could have come closer to something worthy of attention if they had used the data about the journals in which papers were published and had sought data about the rejection rates of the journals involved, data that journals often do publicize.

That they present data on articles published in two distinguished journals suggests that the first part of the data requirement was available. The second data requirement would not have been a major obstacle.

Moreover, I will note that the report indicates that the majority of papers from Taiwan were on “computer science, engineering, economics and business administration” — fields that are not usually of interest to journals such as Nature and Science.

That suggests that the analysis should be done by field of endeavor, rather than lumping everything together. If it had been done on that kind of a basis it might be possible to infer trends more clearly.

Emilio Venezian

New Taipei City

Bad behavior unpunished

Andres Chang (Letter, March 25, page 6) responded to the implicit racism in Pai Po-hsueh’s letter (Letter, March 18, page 8), in which Hsueh demanded that foreigners behave better. However, what makes Hsueh’s complaint particularly ridiculous is the inconsiderate behavior of a minority of Taiwanese people who misbehave on a daily basis (Letter, Oct. 6, page 8).

Comments will be moderated. Keep comments relevant to the article. Remarks containing abusive and obscene language, personal attacks of any kind or promotion will be removed and the user banned. Final decision will be at the discretion of the Taipei Times.

TOP top