A new project has begun. A milkshake bar has opened in town with all kinds of messed-up flavours simply begging to be converted into fat and stored on my thighs. A full survey must be made of their myriad treats. Since I have (optimistically) 18 months left in the North East, I calculate I must sample around ten shakes a month, so perhaps two a week, until I either reach the bottom of the barrel (parma violet flavour isn't something I'm particularly looking forward to), or I get too fat to fit into my office, which most closely resembles a walk-in closet with all the clothes replaced by a quantum theorist with an obsession with bagpipe music.
The up-side of all of this is that I get to do stats. Yay for stats. Each shake will receive a rating in the following categories: taste, texture, synergy (the degree to which the chosen flavour succeeds within the context of a chilled dairy-based drink), and an overall score. General comments will also be provided. Week by week this will build into an indispensable guide to Shakeaholic milkshakes. This guide cannot be purchased in shops!
Today's shake: Maple Syrup
Since this is my first shake, I am presented with something of a dilemma. How can I rate the shake without reference to the others? Should I simply place it in the middle of each category? What if further experience proves that it is one of the poorer shakes? What if no other shake can measure up to its syrupy goodness? Very roughly speaking, this crisis is the central premise of non-parametric statistics; the attempt to extrapolate future events with very little data, and without even really knowing what the fuck is going on at all. To give Dr F's favourite example, if you go to, say, Gabon, and see three people, you can extrapolate the average height of a Gabonese person. You may be fairly far out, but you at least have data that you can relate to in terms of what you know about human physiology in general.
Now imagine you're on Mars and run into three Martians. One is two feet tall, one is eight feet, the next is eleven feet. What can you say about the average height of Martians? You've never even seen one before, you have no frame of reference.
Those who read the preceding paragraph and conclude that there just isn't anything useful to say without more data are shameful quitters. Much of modern statistics is based on taking almost no information and deciding how one can use it to make the very least outrageous bullshit call possible.
In any event, in honour of Doctor F, the maple syrup shake will receive a five in every category, since the only milkshakes I can compare them to are those sold by McDonalds, and I don't think we can really put them on the same scale.
Taste: 5
Texture: 5
Syngergy: 5
Overall: 5
General comments: It is difficult to verify the syrup in question as being particularly maple-esque, but the drink is smooth and sweet, and thus can be considered a good baseline shake.
Update: J-Dog suggests in an e-mail that I might want to add a "Scorn" statistic, demonstrating the degree to which the combination is an outrage against the natural order. It's a good idea, so, unsurprisingly, we shall give the maple syrup shake Scorn: 5.
3 comments:
Not being any kind of expert, or even journeyman in relation to statistics – I would just state that there is probably a more formal way of saying this, but I’ll give it a go.
Surely you take your first Milkshake – establish it as your base (I would just use 0 but you can start at 5 if you really want). Then rate using your criteria off that base (so each criteria gets say +1 for slightly better -1 for slightly worse, with no limit to how high or low you are entitled to put it, just as long as your consistant ). Then when you’ve gathered enough for a reasonable population you can then establish where you want your average grade to lie (so if out of 10 you want say 5 to be average), you can then calculate where all of your milkshake grades reside on the basis of that scale (so if out of 10 in this example you see what your range is for positive and negative and distribute them proportionally so they all fall in the 0-10 range) and give them all the correct rating values…
Of course, Squid, you're here subscribing to the Bayesian Dogma of Precision [Walley, 1991]. See the light and accept you can do things quite happily without weak orderings for preferences [BigHead & Waffles, 2008].
I'd not got around to giving your co-writer his nickname, but I endorse "Waffles" wholeheartedly.
Post a Comment