Calculating P values and 95% confidence intervals from data presented

C. Gissane
2014 Physiotherapy Practice and Research  
Readers of research articles should be in a position to make maximum use of data presented to them. Evidence-based practice directs that treatments be substantiated by research evidence [1] . So whether readers are practitioners or researchers, a basic understanding of statistics is important [2] . Applying the data from research reports will be enhanced if both exact P values [3-5] and confidence intervals for differences are reported [5, 6] . Together, they provide complimentary information
more » ... ] which assists the reader and avoids misinterpreting findings [7] . If either the confidence interval (CI) or the P values are not reported, the reader could calculate them [6] if they possess the correct skills and tools. The tools to compute CIs and P values are provided in many statistical packages such as SPSS, Minitab and R. Unfortunately, once an article has been published the data is in summary form, a reader will not have access to the data from which these statistics are calculated. In spite of requests from authors and editors [2, 5, 8, 9] information such as CIs are not always presented in research papers, but they could be calculated [6] . Similarly, P values are not always reported exactly, but as inequalities (P < 0.05) [10] . The Cochrane collaboration [11] describes how data can be extracted from research reports so that it can be pooled with data from other papers into a single combined estimate for a meta-analysis. Similarly, Hozo [12] offers a technique for computing a mean and a standard deviation from a reported median and range. There is also a specialist
doi:10.3233/ppr-140045 fatcat:kahutia3mbggtg4twypf5wo7ue