Medical Research: Multiple Causes Of False

Over the years, numerous claimed scientific findings have been proven false by researchers who tried to replicate them, or to test their feasibility. In fact, it appears that claimed scientific findings are often just the result of chance or biases; no true relationship existing between the studied variables. The article explores the multiple causes of false positive findings, from sampling, to data analysis through research design and more. The author found that most claimed research findings are false because the majority of them are based on a single study. Major causes of false findings are biases that can be noticed from sampling, data collection, data analysis and even shifting of the initial hypothesis once the data are obtained. These

In fact, he believes that most medical research is flawed because of the way they are designed and conducted, and the data analysis and interpretation. In fact, in most of these researches, conclusions are made from any relationship reaching formal statistical significance. Researchers often only look for correlation between parameters rather than causality. The author expresses a concern on how conflicts of interests introduces bias in medical research. Researchers, often pressured by their employers for results, or in competition with other teams working on the same topic, will either work with a small sample or ignore undesirable correlations to focus solely on the ones that serve their goals. In some fields it has been proven that nothing was truly discovered and that all the false findings were only the results of prevailing bias that was introduced by the pioneers. By conducting a series of test to calculate the probability for a research finding to be true under certain conditions, the author found that there are fields where the chance to have true findings was very low, close to null, but still numerous studies in those fields were

We have learned that data should be handle with care and we need to learn how the data were collected and analyzed before giving credit to the findings claimed based on them. This aligns exactly with what the author is talking about when he says that most of the research findings are false. We also learned that correlation does not always means causation and once again, we find it expressed in this article since most of those research that prove to be false have established a correlation between two parameters even though there is no causation. We have talked about bias and how they affect statistics; this article not only point out the impact of bias of research findings, but also gives a number of reasons leading researchers to introduce, sometimes voluntarily biases in their work. The author talked about statistical significance in which the only criteria was the p-value and this remembered me of the article on “p-hacking” that was part of our readings and it made more sense to