Discrete Data: Measurements done in health and other research are generally of two types, continuous and discrete. For example, blood pressure, body temperature and weight are measured on a continuous scale. But severity of cancer (stages 0,1,2,3,4), intensity of pain (none, low, moderate and high) and degree of depression (none, low, moderate and high) are recorded on a discrete scale.
My main book on statistics, based on my doctoral thesis and subsequent research, deals with methods of analyzing multivariate discrete data when the sample size is small, or moderately large, but the data are skewed. Generally, such methods are computationally intensive. Prior to my work, statisticians had devised efficient algorithms for settings with two or three discrete variables. My research provided efficient computational methods for problems with many discrete variables.
For this research, a paper of which I was the main author was awarded the Snedecor Prize by the American Statistical Association and the International Biometric Society in 1989. It is one of the top global prizes in statistics, and I think no other statistician from Africa has received it.
My book has obtained good reviews in major statistical journals and thus far remains the only book on the subject. But it is targeted at a restricted audience, mostly PhD level students and specialist statisticians.
Ethical Encounters in Scientific Research: Over the years, I have faced major problems relating to scientific integrity and ethics on four occasions.
Encounter 1: In 1993, I and a student whose PhD thesis I had supervised submitted a paper to Applied Statistics, one of the journals of the Royal Statistical Society. Our paper dealt with methods of analyzing stratified binary data with many small sized strata. But there were inordinate delays in the review process. Upon a chance encounter, I found out that the editor had sent our paper for review to the president of the only company that was marketing statistical software for analyzing such data. This person, a well-known statistician, happened to be my former PhD supervisor at Harvard. Whatever the case, there was a clear commercial conflict of interest here. In particular, our paper showed that our method was far more efficient than the one incorporated in the company’s product.
But when I wrote a letter of complaint to the editor, he said nothing wrong had occurred. Even my appeal to the senior statistical editor at the Royal Statistical Society did not go far.
Eventually, our paper was published but by this time the software company had brought out a new version of its product with a faster data processing time! I suspect that it had used the method described in our paper for this purpose. But there was no way to prove this.
In 1989, when the company had just been founded, its president had offered me a 10% stake in the company for an investment of just 1,000 US dollars. But I refused to take up the offer as one cannot be an impartial academic and at the same time have a commercial stake in the outcome of one’s scientific research. The company took off very well and its software sold widely. Had I accepted this offer, I would have earned more than a million US dollars by now.
Encounter 2: In the mid-1990s, I wrote a review paper on the use of the Fast Fourier Transform (FFT) to discrete data analysis. The FFT is a very efficient computational method that widely used in science. It application to discrete data was pioneered by a well-known Harvard professor, and for nearly two decades, he and his students had published many papers on this topic. They had basically monopolized the field. However, I felt that this particular application of the FFT had not been tested thoroughly and journal reviewers had done a poor job in assessing these papers.
My detailed review and comparative evaluation showed that not only was it a computationally inefficient method but also had a marked tendency to generate erroneous results. However, my paper was rapidly rejected by two statistical journals. I was shocked. I suspected that the journal editors were sending my paper for review to the very authors I was critiquing. A clear intellectual conflict of interest was at work.
So I sent it to a third journal. The response was the same—rejection. At this stage, I called up the journal editor and told him that I was willing to fly to his city and give an extended seminar to his department to make my case. I urged him to send my paper to other independent reviewers who had expert knowledge of the FFT but had no links with the Harvard group. Fortunately, he accepted by suggestion. And soon afterwards, I got a letter with highly positive reviews that encouraged me to expand my paper! So I wrote it in two parts, and soon both were papers were in print.
A major consequence of my paper was after it came out, papers on the application of FFT to discrete data all but ceased to come out. This incident shows that if you are Harvard professor, you can write scientific rubbish and get away with it.
Encounter 3: My next encounter, connected with corporate censorship in medical publications occurred in relation to a book review I wrote for the journal Statistics in Medicine. This episode is described in detail in chapter 8 of Under-Education in Africa.
Encounter 4: This encounter, which is what you are referring to in your email, arose after I wrote an evaluation of a large scale, influential WHO-sponsored study of preliminary treatment on suspected severe malaria among children in rural Africa. This study was published in The Lancet with an editorial that described it as one of the ten most important studies of the past decade. Later this paper was awarded The British Medical Journal Group’s Paper of the Year award.
Yet, I had been using this paper in my class at MUHAS as an example of flawed medical research. So I together with a professor of parasitology wrote a paper evaluating the design, implementation, data analysis and implementation of the study. Our evaluation showed that this study was highly flawed and its conclusions were biased, of low precision, and not applicable to the conditions existing in rural Africa.
We submitted it to The Lancet. However, the associate editor dealing with our paper was the same person who had been the associate editor for the paper we were critiquing. His approval of the original paper was thereby also implicitly being challenged by our paper. Not surprisingly, he put so many conditions and restrictions for the publication of our paper that if the version he desired appeared in print, all the main shortcomings of the study we had identified would be glossed over and our paper would seem like an unsubstantiated allegation. I appealed to Ombudsman of The Lancet but it went nowhere.
I withdrew the paper and sent it to a journal dealing clinical trials methodology. There too, an element of conflict of interest was present but after much haggling with the editors our paper was published in full with a highly supportive editorial comment from a leading authority on the methodology of clinical trials. The main author of The Lancet paper wrote a short, tepid, evasive response to our paper which was published in the same journal. Interestingly, all of the many co-authors of The Lancet study did not lend their name to the main author’s response.
I have been told that our paper has had a major impact on research in this field and the use of the treatment strategy that was being advocated. Our paper did not take sides on the issue of whether or not this treatment strategy should be used. We only said that the study in question was of such a poor quality that it could not be used to support the use of this treatment strategy and that another high quality study was needed before a decision could be made. The WHO must have spent a couple of million dollars on this large sample size multinational study! What a waste.
Many Minor Work Related Encounters at UCLA and MUHAS: Over the years, now and then I had to face researchers and senior professors bent on misusing statistical methods, violating study protocol and using short cuts to increase their publication list and drive home pre-determined conclusions. This has included students cheating in exams and research. At both places, I stood up for what I thought was right but had to face unpopularity or worse.