At www.spectator.co.uk/2016/10/how-many-scientific-papers-just-arent-true/ … the link is provided by Robert Farrar. It seems that Robert Horton, editor of the Lancet, is on record as saying 'that much of the scientific literature, perhaps half, may simply be untrue …' and he added, 'science has taken a turn towards darkness.' The article also provides links to reports in pdf format.
Medical research, psychology, and economics are it seems the main culprits visible to Robert Horton – but it probably affects a lot of other science research too as computer simulation is being used instead of hands on experiments (in laboratories) or investigations in the field. Computer simulation has its place – and some good research is able to exploit that method (but should be backed up by other evidences). Computers can only juggle with the information fed into them and therefore cannot amount to solid evidence (but sometimes it is presented as such). Computers can infer rather than establish. It opens up avenues that can be further explored. However, building a model on a computer simulation is sometimes followed by further research that creates a new model based on the older simulation – and that is airie fairie. This is one of the problems with climate models. It is self delusion masquerading as objective science.
Most people regard the latest medical revelation a pinch of salt, don't take a lot of notice of what psychologists might say, and wouldn't place too much faith in your average economist. In that sense Robert Horton has not said anything outrageous.
Back in February 2016 Nature had an article with the title, 'Mistakes in peer reviewed papers are easy to spot but hard to fix' which probably has it down to a tee.