Wednesday, March 28, 2007

The statistics of mood change

Hello all

The clocks went forward on Sunday morning, and by the end of Monday I had noticed a lifting of mood amongst the Institute staff and visitors. I put it down to the increased light availability and higher doses of vitamin-D, but now I'm not so sure, as other explanations have come to, er, light. As a scientist, I am always interested in the effect of these other explanations (aka CONFOUNDING FACTORS). If they are not taken into account, we all run the risk of BIAS in our INFERENCES and CONCLUSIONS.

In the case of the mood-lightening, I think there are two confounders...
1) the Annual pay rise which came into effect on Monday at 2pm
2) the arrival of our new receptionist, Jemima, at 1pm on Monday.

Items 1 and 2 are confounders because they were both associated with the clocks going forward, and are likely candidates for increasing mood. The trick, and this is a general principle in these matters, is to adjust for the confounders in one's analysis. If the correct adjustment is made, the truth will emerge. In the present analysis, I wanted to ascertain whether it was the clocks going forward, the pay rise, or Jemima that was causing the mood change. At the same time, I was aware that a combination of any two factors could MODIFY the degree of mood change, and I was therefore keen to investigate evidence of INTERACTION between all three factors.

The table below corresponds to the response of everyone in the Insitute at 5pm on Monday when I asked them about each of the above changes






NamePosnClocksRaiseJemima
RavelR.A.I forgotIt is not so muchShe look good, eh boss?
JennyCleanerSo light!Was that a raise? I think her roots need attention
DoloresWife Don't careI'm surprised you still have staffWas she the only candidate?
DaveVisitorSorry I was late this morning!N/APhew!
AngelaVisitorIs it spring forward, fall back or the other way round?N/AWhat happened to Denise?

Upon visually inspecting the table, I was left wondering how to analyse the data. The SAMPLE SIZE was particularly small, and there was a wide variety of responses. Standard statistical analyses were not possible, so I relied on a little known technique of EYEBALLING the data.

My visual anlaysis revealed that there was no clear winner in terms of the main EXPLANATORY FACTOR. Most individuals to whom the pay rise was applicable seemed content, and there were mixed repsonses, though generally positive, towards Jemima. Dolores was the only respondent to have persistently negative responses, and I am therefore going to undertake a POST-HOC analysis of her in isolation.

CONCLUSIONS
The increase in mood throughout the Institute (excluding Dolores) was due to a combination of increased light levels, the arrival of new member of staff, and the annual pay rise. No one factor dominated.

J McC

No comments: