P-values extracted from 11 scientific journals across two years

We examined how the proportion of reported significant results, expressed as p-values, changed with increasing impact factor (IF) We divided the range of IF into three intervals: low (IF<4), medium (4≤IF<8), and high (IF≥8) to ensure that our model was applicable to a broad range of impact factors. We then randomly selected at least two journals for each IF interval for both 2012 and 2014 from journals listed in the Science Citation Index Expanded (http://mjl.clarivate.com/cgi-bin/jrnlst/jloptions.cgi?PC=D) under four subject categories: biodiversity conservation, biology, ecology, and evolutionary biology. We examined 11 journals (lowest IF 0.36 to highest IF 17.95), six journals for both years, whilst we examined three for 2014 and two for 2012 alone. We collected p-values reported in all articles using the free trial version of Examine32 Text Search from Aquila Software. We extracted exact p-values and inexact p-values (e.g. p<0.05). All the inexact p-values were reassigned to six intervals in order to calculate the proportion of significant results (Table A in S1 Text). Due to the ubiquitous use of 0.05 as the alpha level in ecology, we considered p-values below this level as significant.