Sunday, 9 March 2014

A Tale Of Cocktails: Facts, We Haz Them

Been a while since we did this, so let's take another stupidly long look at the various ways I've been showing my exocrine system who's boss.


11 Best Cocktails

1. Brain Hemorrhage
2. Flying Grasshopper
3. Woo Woo
4. Fuzzy Shark
5. Choc Berry
=6. Baby Guinness
=6. Dennis the Menace
8. Sugarific Ciderific
=9. Malibu Pop
=9. Daiquiri

5 Worst Cocktails

1. Screwdriver
2. Champagne Cocktail
3. Orange Blossom
4. Tomorrow We Sail
5. Poinsettia Holiday

9 Tastiest Cocktails  

1. Woo Woo
=2. Flying Grasshopper
=2. Midori Sour
=2. Choc Berry
=2. After Six
=2. Baby Guinness
=2. Brain Hemorrhage
=2. Mudslide
=2. Malibu Pop  

Worst Tasting Cocktail


6 Prettiest Cocktails  

1. Brain Hemorrhage
=2. Metropolitan  
=2. Midori Sour
=2. Choc Berry
=2. Fuzzy Shark
=2. Baby Guinness  

Ugliest Cocktail


11 Cheapest Cocktails  

= Caribbean Milk
= Ciderific
= Sugarific Ciderific
= Screwdriver
= Daiquiri
= Raspberry Tipple Plus
= More Sunshine
= Dribena
= Fuzzy Shark
= Fuzzy Navel
= Snowball  

Most Expensive Cocktail


8 Best Named Cocktails  

=1. Woo Woo
=1. Brain Hemorrhage
=2. Flying Grasshopper
=2. Metropolitan
=2. Daiquiri
=2. Choc Berry
=2. Fuzzy Shark
=2. French 75

2 Worst Named Cocktails  

= Tomorrow We Sail
= Champagne Cocktail  

12 Easiest Cocktails  

=1. Elderflower Royale 
=1. Kir Imperial
=3. Sex on the Beach
=3. Flying Grasshopper
=3. Dribena
=3. Black Forest
=3. Ume Royale
=3. Kir Royale
=3. Blue Lagoon
=3. Baby Guinness
=3. Brain Hemorrhage
=3. Mimosa  

Most Fiddly Cocktail  


11 Strongest Cocktails  

1. Flying Grasshopper
=2. Baileys Cookie Martini
=2. Caipirinha
=2. White Lady
=3. Brain Hemorrhage
=3. Mudslide
=5. After Six
=5. Dennis The Menace
=5. Baby Guinness
=5. Malibu Pop
=5. Champagne Cocktail

Weakest Cocktail

Choc Berry

Supplies Consumed


Blue Curacao
Cherry wine
Chocolate liqueur
Creme de Cacao
Creme de Cassis
Creme de Menthe
Elderflower liqueur
Irish Cream (Baileys)
Peach Schnapps
Plum wine
Rum (dark)
Rum (white)
Sloe gin
Tia Maria
Triple Sec
Vodka (standard)
Vodka (vanilla)


Cranberry juice
Lemon juice
Lime cordial
Lime juice
Orange juice
Pineapple juice
Sugar syrup
Tonic water
Vanilla syrup


Caster sugar
Mint Matchmaker
Orange slice
Orange peel
Whipped cream
A shark



Estimated amount of ice used: 768 cubic centimetres.


Mean cocktail score: 6.75

Standard deviation of cocktail score: 0.777

Range of scores: 4.2

Last time around we demonstrated a rough normal distribution of the cocktail scores.  But that was just lazily eyeballing the QQ plot.  This time let's get serious and run them through a Shapiro-Wilks test, the result of which is a p-value of 0.099.  Whew!

With that done, we can fit a linear regression, with the overall scores as output and the various booze types as the covariates.  This way we may be able to detect what ingredients have a significant effect upon my enjoyment or otherwise of mixed drinks.  45 cocktails is admittedly a fairly small sample size to subject to a linear analysis, particularly with so many covariates, but we'll see what happens.

Eventually we'll want to consider combinations of drinks, but for now lets work through each booze type separately.  The first number is the  p-value; a p-value of 0.05 or less can be (tentatively) be considered significant (in layman's terms; it looks like the ingredient has an effect on my tastes above what might be accounted for by chance). The attached number tells us the average effect of the booze type upon cocktail score.

(As an example, then, Baileys might look like it has a larger effect on my opinion than champagne - and a positive one, since adding Baileys to a cocktail increases its score by 0.683 on average, but unlike champagne, we cannot say this effect is not down to chance. The reason we can't say that for Baileys even though the apparent effect is larger than that for champagne is that I've had fewer cocktails containing Baileys than including champagne, which makes it harder to pin down the effect.)

Advocaat 0.562 -0.464
Amaretto 0.309 -0.579
Baileys 0.064 0.683
Blue Curacao 0.422 0.300
Brandy 0.058 -1.49
Cachaca 0.651 -0.361 
Champagne 0.027 -0.573
Chambord 0.697 0.161
Cherry Wine 0.651 -0.361
Chocolate Liqueur 0.459 -0.422
Cider 0.258 0.652 
Creme de Cacao 0.043 1.58
Creme de Cassis 0.788 0.154
Creme de Menthe 0.078 0.991
Dark Rum 0.836 -0.086
Elderflower Liqueur 0.272 -0.873
Gin 0.236 -0.408
Kahlua 0.556 0.243
Malibu 0.458 0.278
Midori 0.952 0.048
Peach Schnapps 0.020 0.695
Plum Wine 0.952 0.048
Port 0.174 -1.08
Tia Maria 0.337 0.764
Triple Sec 0.697 0.161
Vanilla Vodka 0.952 0.048
Vodka 0.960 0.014
White Rum 0.483 0.559

So apparently cocktails with creme de cacao or peach Schnapps in them are significantly better, and those with champagne are significantly worse.  Of course, just saying such a thing reveals the problem here: there are plenty of cocktails adding Schnapps to probably cause problems (something involving creme de menthe, for instance, which itself came close to being significant and which apparently adds a full point on to each cocktail into which its placed), and there are some cocktails so hideous that a burst of bubbly would be helpful, if only to dilute the horror.

Ultimately what's really needed here is an analysis of the covariances.  That will have to wait until I've consumed more cocktails before it seems worth doing, however.  To the magic cupboard!

No comments: