This tweet of Ben Goldacre’s caught our attention yesterday:
 ben goldacre (@bengoldacre)17/01/2011 13:59 sigh. #bluemonday is trending. but #bluemonday is PR bullshit. humour me, RT and flood the tag. http://bit.ly/i21KLt |
His linked blog post has more details: do have a look.
So, what do the mappiness data tell us about people’s happiness yesterday on ‘Blue Monday’, 17 January 2011?
First, a simple bar chart showing mean happiness responses, by day, since mid-September. Weekends are in green. Yesterday is in yellow, on the right, and a horizontal line is drawn across from that bar for comparison.
Nothing particularly special about ‘Blue Monday’ is apparent here. What mainly leaps out is just how happy the Christmas season is, including a huge upward spike on Christmas Day.
The time trend could, however, be confounded by the fact that our participants now are largely not the same people as our participants were in September: it’s plausible that happier or less happy people could have joined up at different times.
The time trend could also be confounded if there’s a trend in each individual’s responses, and if our average participant now is at a different stage of participation than the average participant was in September.
The chart below, therefore, removes variation between participants, and between different stages of participation, from the numbers being plotted. The scale down the left also changes, but since we’re interested in relative effects this isn’t really important. (The chart shows the residuals from an individual fixed-effects regression, which included dummy variables for various stages of participation. We added a positive constant to the residuals to keep the chart’s shape vaguely comparable to the one above).
These corrections do make ‘Blue Monday’ look a little worse. But there are still several less happy Mondays spread throughout the preceding months, and an even greater number of less happy Tuesdays.
So, for our participants at least, it seems ‘Blue Monday’ really is just PR BS.