There has been a bunch of research on the likelihood of bad things happening on Friday the 13th. These researchers thought that the big problem with all of the available research is that the populations studied have been too small. The authors took information on over 56,000 patients with acute coronary syndromes, broke them down into 217 day/date combinations (Friday the 1st, Saturday the 1st, . . . ,Wednesday the 31st, Thursday the 31st), and compared the outcomes of those 216 groups with their Friday the 13th group.
Cut to the conclusion –
Conclusion: On most days, there was no difference in the 13-year mortality rate for patients admitted with their first ACS from that for “unlucky” Friday the 13th. However, patients admitted on five day/number combinations were 20-30% more likely to survive at 13 years. These findings could be explained by subgroup analysis inflation of the type I error, although supernatural causes merit further investigation.
No. Supernatural causes do not merit further investigation, at least, not based on anything in this paper.
The authors used Friday the 13th as their normal date for comparison with every other date, but the outcomes from Friday the 13th are not the true statistical mean. The outcomes on Friday the 13th were just chosen because of the superstition being investigated. Friday the 13th is so close to the statistical mean that this mistake is easy to make.
Surprisingly, however, we also identified five potentially “lucky” days on which mortality rates were significantly lower, by 20-30%.
When analyzing 217 samples, it is not surprising that some of the data deviate from average by an amount that is expected to produce no more than one significant deviation out of every twenty comparisons. The authors had over 200 comparisons, so we should not have been surprised by up to 11 day/date combinations with p values of less than 0.05. There were only 5. Should anyone go looking for supernatural explanations for statistically normal outcomes?
While Friday the 13th was not the statistical mean, it was very close. Look at the five potentially “lucky” days and how close the ranges are to 1.00. If the range crosses (includes) 1.00, the results are not statistically significant according to the prospectively determined criteria of the authors. Crossing 1.00 is just another way of expressing P <0.05. Sunday the 1st and Monday the 29th each produced outcomes 29% worse than Friday the 13th. Saturday the 31st produced outcomes that were 36% worse. If we compared these with the actual statistical mean, Monday the 29th and Saturday the 31st become significantly “unlucky” using a p value of less than 0.05 and all of the significantly “lucky” days become insignificant.
As we should expect, the most extreme benefit and harm both fall on the 31st. Only 7/12 (58.3%) of months have 31 days, so these days have much smaller sample sizes. With smaller samples, the appearance of deviance is expected to be greater. The actual deviation is less important, because the sample size is smaller.
Friday the 13th is only slightly different from the statistical mean, using the data in this paper, which may be the largest examination of a possible Friday the 13th effect.
Once again, the biggest problem with Friday the 13th is that we end up listening to people promoting superstition.
I have also written about this kind of superstition here –
The Magical Nonsense of Friday the 13th – Fri, 13 May 2016
Happy Friday the 13th – New and Improved with Space Debris – Fri, 13 Nov 2015
Blue Moon 2012 – Except parts of Oceanea – Fri, 31 Aug 2012
2009’s Top Threat To Science In Medicine – Fri, 01 Jan 2010
T G I Friday the 13th – Fri, 13 Nov 2009
Happy Equinox! – Thu, 20 Mar 2008
Protty, M., Jaafar, M., Hannoodee, S., & Freeman, P. (2016). Acute coronary syndrome on Friday the 13th: a case for re-organising services? The Medical Journal of Australia, 205 (11), 523-525 DOI: 10.5694/mja16.00870