Friday, 24 February 2017

Replication Across the Science Nation?


Some interesting findings from the attempt to replicate findings from some cancer research projects (www.nature.com/news/cancer-reproducibility-project-releases-first-results-1.21304). The study started in 2013 and will attempt to replicate some 29 studies in the general area of cancer research. To date, the results of attempts to replicate 5 studies have been released with 1 failure to replicate, 2 partial replications and 2 studies which are difficult to interpret. The first thing to say, is that replication is a necessary element in the scientific method (although I have been asked in the past why it is necessary to repeat tests with more animals by people concerned with animal use). The difficulty is that science deals in probabilities rather than absolutes. Further problems seem related to a) the pressure to publish positive results (negative findings are difficult to publish); b) media desire for positive stories of 'cures' and 'major breakthroughs' and c) the need to encourage further funding of programmes (things that also have a direct impact on the individual's perceived worth by institutions and universities). These are all factors that encourage early (premature?) exposure of results. One must also note that scientists are often somewhat unskilled in statistical methods. I am certain that these issues are not limited to cancer research.

No comments:

Wooden Tops 20. Horse chestnut

Horse chestnut wood is rather weak but has been used to make furniture. Its absorbant properties, however, make it ideal for fruit racks an...