The Rosie Project Alice Richardson, Maths & Stats, ESTeM.

The Rosie Project

Alice Richardson, Maths & Stats, ESTeM

• Simsion, G. (2013). The Rosie project. Simon & Schuster.

Statistical terms Don bandies about

• Chapter 1 “Julie, the convenor, … blonde with big tits. In fact, her breasts ere probably no more than one and a half standard deviations from the mean size for her body weight, and hardly a remarkable identifying feature.”

• Standard deviation measures spread in a variable e.g. breast size

• It is most useful when the variable follows a Normal distribution

• If breast size is Normally distributed, then about 2/3 of women are within one standard deviation of the mean

-3 -2 -1 0 1 2 3

0.0

0.1

0.2

0.3

0.4

standardised breast size

de

nsi

ty

68%

• And about 85% are within 1.5 standard deviations of the mean

-3 -2 -1 0 1 2 3

0.0

0.1

0.2

0.3

0.4

standardised breast size

de

nsi

ty

85%

• Chapter 4 “… best practice in questionnaire design, including multiple-choice questions, Likert scales, cross-validation, dummy questions and surrogates.”

• Multiple choice questions: select from several options rather than write in an answer or have only two options.

• For an appointment, do you arrive

• (a) very early• (b) a little early• (c) in time• (d) a little late• (e) very late

• Likert scales scores responses along a symmetric range

• Likert, R. (1932). A technique for the measurement of attitudes. Archives of Psychology 140.

• For an appointment, do you arrive

• (a) very early• (b) a little early• (c) in time• (d) a little late• (e) very late

• Cross-validation • “Height, weight and body mass index. Cant you do the calculation yourself?”

• “That’s the purpose of the question. Checking they can do basic arithmetic.”

• Dummy questions • Questions not used to calculate final scores but to construct other variables e.g. SES or to calibrate other questionnaires

• Surrogate questions• “Q35: Do you eat

kidneys? Correct answer is (c) occasionally. If you ask directly about food preferences, they say “I eat anything” and then you discover they’re vegetarian.”

• Chapter 4 “My strategy was to minimise the chance of making a type-one error – wasting time on an unsuitable choice. ”

• H0: this person will not make a suitable wife

• Ha: this person will make a suitable wife

• Reject H0 wrongly = Type I error

• Fail to reject H0 wrongly = Type II error

• Chapter 4 Inevitably that increased the risk of a type-two error – rejecting a suitable person.”

• Minimise Type I error completely by NEVER getting married = high risk of Type II error (failing to marry a suitable person)

• Reject H0 wrongly = Type I error

• Fail to reject H0 wrongly = Type II error

Conclusion

• A little statistics goes a long way!

• Hofstadter , D. (1979). “Godel, Escher, Bach: An eternal golden braid”

• A metaphorical fugue on minds and machines in the spirit of Lewis Carroll