Is blogging better than publishing?

This week I started with the intension of writing about the use of Mata for analysing the Poisson regression problem that has occupied this blog for the last month. Before preparing the Mata program, I went back to my Stata code to use it as a guide and when I did so, I noticed that in going from the slow Stata program based on the logdensity function to the faster Stata program using my purpose-written code for the log-densities, I had left out the priors. I thought to my self,

‘Oh well, these things happen, I’ll put it right in my next posting’.

Later I checked and found that I was mistaken and in fact I had not forgotten the priors at all but, by the time I had realized this, a thought had already occurred to me,

if such an error had occurred in a published paper I would have been very embarrassed and I would have felt obliged to have published a correction (though I wonder how many researchers actually keep quiet about such errors and hope they go unnoticed)

So today I want to muse on this point. Next time I’ll return to writing Mata code.

Errors, such as leaving off a term in a long calculation, occur all of the time. I even teach my students that

‘it is not the errors that you make that are important, it is whether or not you spot them’

So in my teaching I emphasise careful checking of the results to make sure they agree with what was anticipated and if they don’t, we try to find out why.

In my blog I would feel quite happy to admit such an error, afterall, in reality, that is what happens in all complex statistical analyses. Yet I know that the same error in a published paper would give me sleepless nights. So, are there ways in which blogging is better than publishing?

Perhaps the potential advantages of blogging about research stem from the freedom to be more honest and to show the process of arriving at an answer, including the things that don’t work and the errors that are made on the way.

Academic publishing has been under pressure for some time. Publishing a journal is expensive and as the cost of journals rises, so libraries find it increasingly difficult to afford the subscriptions. Add to this the huge expansion in the amount of research that is undertaken worldwide and the fact that every year new journals are started and it is clear than the existing financial model is under threat. In response to this pressure, attempts have been made to make the author pay and to make the journals free to readers. Such a change raises the cost of doing research and discriminates again researchers who do not work in well-funded institutions in rich countries; it also places subtle pressure on the journals to accept papers of lower quality when the mere act of publishing brings in money.

The advent of electronic journals has placed even more pressure on traditional publishing. I cannot remember the last time when I went to my University’s library to look at a paper copy of an article; it must be many years ago. Now all of my work is digital. Even the printed journals are all accessible electronically and this format is far more convenient.

There is a saying in my line of work, that if all else fails you can send your paper to PLoS one. This might sound slightly insulting but actually I think just the opposite. PLoS one is an electronic journal that has a very broad coverage and a very generous acceptance policy. Here is a quote from its own website,

“All good science deserves to be published.

PLOS ONE takes the hard work out of publishing. There’s no stress waiting to find out if your article meets subjective acceptance criteria. As long as your work reaches a high technical and ethical standard, PLOS ONE will publish it – and make it freely available to a global audience.”

So for PLoS one, provided that the paper describes original, ethical research to a good academic standard, it will be accepted. Papers are still peer reviewed but there is no assessment as to whether the paper is ‘important’. This seems to me to be an excellent basis for publication, although there is still a deal of resistance to it. If you want to be promoted in a British university then it is much better to publish in a highly selective journal like Nature Genetics rather than an unselective journal such as PLoS one. People are judged not on what they do, but on where their work is published. One can sympathise with this approach when a University wants a quick and easy metric of performance, but that does not make it a good system and it does not mean that it is system with a long future.

Now let me return to blogging and ask what role it has in research. The limitation of journals, even inclusive ones like PLoS one, is that they only publish the final article. They do not show the process that went on in order to get there. So the errors are (hopefully) removed, the attempts that failed are ignored and, as a statistician, I often find it hard to work out why one method of analysis was adopted while others were rejected. Are we merely being told about the analysis that gave the ‘best’ results? where best might mean most significant or most in keeping with some particular theory.

Could a blog fill in the gaps? Suppose that researchers were in the habit of opening a blog whenever they started a new research project. They could comment on progress on a regular basis, including failed experiments and abandoned theories and unsatisfactory analyses. Being electronic they could even release preliminary data or early results. Hopefully the blog would chart the progress towards the final publications.

Being public, not only would readers gain from seeing the truth about the process of research but the authors would gain from feedback from the readers; potentially a real force for improving the quality of the research. It would also mean that there would be a record of research projects that do not make it to final publication. Much is made in my own area of research of publication bias whereby a proportion of studies, usually those with negative or puzzling findings, do not get published. A research blog would help address this issue.

I can imagine that many people will find this idea very challenging; they will have been brought up in a world in which ideas are kept secret until the last moment so that they cannot be stolen by another researcher and data are kept private so that other researchers cannot take advantage of your hard work. I think that it is clear that these attitudes are based on personal gain and not on the best interests of science. They also reflect the myth behind the Nobel prize, which is that science advances because of the efforts of a few great minds, when in reality most good ideas are shared by many people and are worked on by large groups before, often arbitrarily, one senior person takes the credit.

Now if we accept the blog as an electronic record of the progress of a research project, then why do we still need journals? Why not allow the project to publish its own final reports as part of the blog?

It seems to me that the journals offer two advantages; peer review and discipline over the length and format of the report. It is true that journals also used to have a role in collating together work in a single research area but in an electronic age people no longer search by journal. As to peer review, surely we could organise a central database of reviewers willing to comment on each other’s work, essentially this is all that journals do. So researchers could use the database to arrange their own peer review, although they would need to publish the names of the reviewers in order to guard against fraud (and why not publish the reviews?). The system might seem more likely to be abused but this implies a comparison with the myth that currently published work is honest and error free.

Discipline over length and style is important, although in an electronic age journals themselves are moving further and further from this aim by publishing longer and longer supplements without any specified format. Recently, I agreed to review an article only to discovery that it came with an 80 page supplement. We could easily devise a standard for publication of a final report, like a standard for conducting a clinical trial or a standard for medical ethnics. Then researchers could say that their final report is produced subject to a particular standard.

I have to acknowledge that I have painted a one-sided and rather idealistic picture, I do not really believe that academic journals will disappear any time soon, but perhaps an element of blog-like reporting and more openness about the things that go wrong in scientific studies would lead to better science.