Re: ALPSP Research study on academic journal authors

From: Jim Till <till_at_UHNRES.UTORONTO.CA>
Date: Tue, 30 Jan 2001 14:03:35 -0500

On Tue, 30 Jan 2001, Sally Morris wrote:

> I have been asked whether the acceptance/rejection figures varied
> significantly by subject area, so I have delved deeper into the figures to
> analyse this. The provisional results are interesting (bear in mind,
> though, that the samples for some subjects are very small)
>
> By and large, the arts and humanities journals (if I may call them that)
> appear to be far fussier than those in the sciences, with a marked skew
> towards a low percentage of acceptances. I attach a table for those who can
> read it.

Sally, thanks for this first look at the data. They are interesting
(interesting enough for me to spend a lunch hour looking at them!).

I've tried to convert the 'percentage responses' into actual numbers
(rather than percentages). I've assumed that the first set of data for
'Life Science' are the correct ones (not the second). The data for
'Medical and Veterinary Science' don't add up to the expected total
(should be 53 responses, not 44?), which accounts for the question marks
in the table below. (I hope that the table can be read!).

For categories involving more than 10 journals:

Percent
acceptance....E&M...LS...M&C...M&VS...SS&E...Total

under 10.......0.....0....0.....1?.....1.......2?
10-25..........0.....1....0.....2?....10......13?
25-50..........0....14....3....14?.....7......38?
50-75..........6.....9....1....21?.....3......40?
over 75........0.....5....0.....6?.....0......11?

Total..........6....29....4....44?....21.....104?

Because many of the cells in this table are small, an exact statistical
analysis isn't easy for an amateur statistician to perform! However, I
combined some of the acceptance rate categories, in order to obtain
multiple 2x2 tables (0-50% acceptance vs 50-100% acceptance), and applied
Fisher's Exact Test. (I still have tables on my bookshelf for this test,
compiled by DJ Finney and colleagues: 'Tables for testing significance in
a 2x2 contingency table', Cambridge University Press, 1963!).

Because of the very small total (4 responses) for the M&C category, and
the uncertainty about the correct numbers for the M&VS category, I did
only two comparisons:

E&M (Engineering and Materials Science) vs LS (Life Science): E&M appeared
to have significantly *more* responses (than expected from the marginal
totals) in the 50-100% acceptance category (P=0.024).

SS&E (Social Science and Education) vs LS (Life Science): SS&E appeared to
have significantly *fewer* responses (than expected) in the 50-100%
acceptance category (0.05>P>0.01).

Even when one takes into account that I did more than one comparison, it
appears that, on the basis of these figures, SS&E journals do have a lower
acceptance rate, and E&M journals a higher acceptance rate, than LS
journals. So, these data do seem to be consistent with the results of the
Zuckerman and Merton study (referred to in previous messages).

Jim Till
University of Toronto

The remainder of Sally's original message was:

> If we look only at those samples covering more than ten journals:
>
> Engineering & Materials Science (12 journals, 6 responses)
> All of the responses showed between 50 and 75 percent acceptance
>
> Life Science (39 journals, 29 responses)
> Under 10 percent - none
> 10-25 - 3 percent
> 25-50 - 46 percent of journals, 48 percent of respondents
> 50-75 - 33 percent of journals, 31 percent of respondents
> over 75 - 18 percent of journals, 17 percent of respondents
>
> Mathematics and Computing (11 journals, 4 responses)
> Under 10 percent - none
> 10-25 - none
> 25-50 - 64 percent of journals, 75 percent of respondents
> 50-75 - 36 percent of journals, 25 percent of respondents
> over 75 - none
>
> Life Science (39 journals, 29 responses)
> 10-25 - 3 percent
> 25-50 - 46 percent of journals, 48 percent of respondents
> 50-75 - 33 percent of journals, 31 percent of respondents
> over 75 - none
>
> Medical and Veterinary Science (66 journals, 53 responses)
> Under 10 percent - 2 percent
> 10-25 - 3 percent
> 25-50 - 24 percent of journals, 26 percent of respondents
> 50-75 - 38 percent of journals, 40 percent of respondents
> over 75 - 15 percent of journals, 11 percent of respondents
>
> Social Science and Education (26 journals, 21 responses)
> Under 10 percent - 4 percent of journals, 5 percent of respondents
> 10-25 - 42 percent of journals, 48 percent of respondents
> 25-50 - 35 percent of journals, 33 percent of respondents
> 50-75 - 19 percent of journals, 14 percent of respondents
> over 75 - none
>
> So insofar as these figures are representative (they cover just over 200
> journals), there does seem to be some bias towards lower average acceptance
> rates (i.e. higher rejection rates) in the arts and humanities than in the
> sciences. What that tells us I am not sure!
>
> Sally
Received on Wed Jan 03 2001 - 19:17:43 GMT