Reporting results and drawing conclusions
Learning outcomes
On watching this video students should be able to:
- Explain that when drawing conclusions from inferential results we often have to exercise good judgement on aspects other than the solitary confidence interval or p-value that we are interpreting; for example, have multiple tests been performed? is the study powerful enough? Is this effect of any practical significance?
- Discuss the aphorism “absence of evidence does not mean evidence of absence” and apply it to exercise good judgement when interpreting larger p-values and wide confidence intervals that contain the null value.
- Justify why two-tailed p-values are generally preferred over one tailed p-values.
- Define a type 1 and type 2 error.
- Explain why a post hoc test can only be interpreted as hypothesis generating and why it is not valid to trawl the data for patterns and then perform hypothesis tests on the patterns you see.
Correction:
07:20: I describe what a type 2 error is incorrectly. The text on the video is correct.
Optional further reading:
Altman DG and Bland MJ. 1995. Absence of evidence is not evidence of absence. BMJ. 311
Gardner MJ and Altman DG. 1986. Confidence intervals rather than p-values: estimation rather than hypothesis testing. BMJ. 292
Goodman S. 2008. A dirty dozen: 12 p-value misconceptions. Seminars in Hematology. 45:135-140
Sedgewick P. 2012. Absolute and relative risks. BMJ. 345