Figure 1.
Positive Results by Discipline.
Name of discipline, abbreviation used throughout the paper, sample size and percentage of “positive” results (i.e. papers that support a tested hypothesis). Classification by discipline was based on the Essential Science Indicators database, the hard/soft, pure/applied and life/non-life categories were based on previous literature (see text for details). Error bars represent 95% logit-derived confidence interval.
Figure 2.
Positive Results by Disciplinary Dimension.
Number of papers that supported (white) or failed to support (black) a tested hypothesis, classified by disciplinary categories based on dimensions identified by previous studies (see text for explanations). Percentage in each bar refers to positive results. OR = Odds Ratio (and 95%Confidence Interval) of reporting a positive result compared to the reference category of Hard/Pure disciplines. Chi square was calculated for each dimension separately (for category composition see Fig. 1).
Figure 3.
Positive Results by Disciplinary Domain.
Percentage of papers that supported a tested hypothesis, classified by disciplinary domain. Blue = including only pure disciplines, Red = including only applied disciplines, Black = all disciplines included. Error bars represent 95% logit-derived confidence interval. For domain composition see Figure 4.
Figure 4.
General Methodology by Discipline and by Domain.
Methodology employed by papers in different disciplines and domains. Methodological categories correspond to basic characteristics of the outcome: whether it measured physical/chemical parameters as opposed to behavioural parameters, and whether the object of study was non-biological, biological non-human, or biological human (see Methods for further details).
Figure 5.
Positive Results by Methodological Category.
Percentage of papers that supported a tested hypothesis in pure (top) and applied (bottom) disciplines, plotted by general characteristics of their methodology (defined by the outcome, see also Fig. 4). The “other methodology” category is not shown. Black = studies on non-human material or subjects, Red = studies on human material or subjects. Error bars represent 95% logit-derived confidence interval.
Table 1.
Logistic regression slope, standard error, Wald test with statistical significance, Odds Ratio and 95% Confidence Interval of the probability for a paper to report a positive result, depending on the following study characteristics: discipline of the journal in which the paper was published, papers testing more than one hypothesis (only the first of which was included in the study).
Table 2.
Logistic regression slope, standard error, Wald test with statistical significance, Odds Ratio and 95% Confidence Interval of the probability for a paper to report a positive result, depending on the following study characteristics: disciplinary domain, papers testing more than one hypothesis (only the first of which was included in the study), and papers published in pure as opposed to applied disciplines.
Table 3.
Logistic regression slope, standard error, Wald test with statistical significance, Odds Ratio and 95% Confidence Interval of the probability for a paper to report a positive result, depending on the following study characteristics: methodological category, papers testing more than one hypothesis (only the first of which was included in the study), and papers published in pure as opposed to applied disciplines.