Probabilities Cheat Sheet

Probabilities... (A.K.A. Chance) Definitions

How likely something (an event) is to happen
Coefficient – Basically just a static numerical value that
is used in a calculation
Kinds of Probabilities
Linear – Like, or in shape of, a line
Conditional Probabilities – Probability of an event
happening based on whether or not something else Variable – Something we take into account in our
happened analysis
Join Probabilities – Probability of two events Remember! Probabilities are always
Ordinal Data – Data that is ordered so that its values
happening at the same time between 0 and 1, if you get a probability
indicate rank
outside this range there is something wrong
Marginal Probabilities – A.K.A. Unconditional with calculation.
Dichotomous Data – Data that takes on two values only
Probabilities, are just the summation of all (e.g. 1 or 0, True or False, Yes or No)
probabilities 0 = Not gonna happen
1 = Definitely gonna happen
How many times event happened
Probability =
Total Outcomes Big Takeaway Logical Fallacies to Avoid...
Probabilities can give you an indication of what is ...to make better arguments
Kinds of Events likely to happen, but they CANNOT tell you what
will happen. Fallacy – An error in reasoning that will undermine your
Mutually Exclusive – Events that can’t happen at
argument
same time Here’s an example… probability tells us that when
you flip a coin you have a 50% chance of hitting Slippery Slope – Saying that if A happens, and B-Y
Non-Mutually Exclusive – Events that can happen at either heads or tails, so if you flip the coin 100 times happen, then Z will happen, so basically A = Z
same time you would expect to get 50 heads and 50 tails… but
that isn’t always what will happen! Hasty Generalizations – Just what it sounds like, you
Independent – When an event’s probability isn’t
come to a conclusion about something before you have
affected by anything else happening or not happening All that probability is telling us is that most of the sufficient information
(e.g. a coin toss isn’t affected by previous coin toss) time you’ll get a number somewhere close to 50
heads and 50 tails, not that you absolutely will get Post hoc ergo propter hoc – Basically saying that if B
Dependent – Events whose probabilities change that exact number each and every time you run the happens after A, then A caused B
based on each other happening or not happening “100 coin flip test.”
Genetic – Saying that the origin of a person or a thing
dictates its character or worth

Begging the Question - When you try to validate your

Correlation conclusion within the question that you are asking
When there is some relationship between two things
Circular Arguments – Stating a conclusion that just
Correlations always take values between -1 and 1 restates itself as proof (ex. My car is awesome because
it’s so cool)
-1 is a perfect negative correlation, which means as one thing gets bigger the other thing gets smaller
Either-Or – Oversimplifying a conclusion by assuming
0 is no correlation at all, basically there is no relationship between these things that it must be either one thing or the other
1 is a perfect positive correlation, which means that when one thing gets bigger so does the other

The closer the correlation value is to -1 or 1, the tighter (more linear) the relationship will be on a scatter plot
(see below on Pearson’s coefficient)
Caution Hazard
Perfect High Low
Positive Positive Positve No
Correlation Correlation Correlation Correlation Correlation DOES NOT prove
Causation!
Beware temptation to say that a correlation
between two things means one causes the other.

For example…
There may be a correlation between sweater and
1 0.9 0.5 0 snow-shovel sales. However, that does not mean
that sweaters make people buy snow-shovels.
Low High Perfect
Negative Negative Negative All that we can say with a correlation is that there is
Correlation Correlation Correlation a relationship/link between sweater sales and
snow-shovel sales.

Multiple & Conditional Probability

Make sure that you account for ALL possible events
when calculating probability.

-0.5 -0.9 -1 Same is true for conditional vs. unconditional

probabilities, be sure you understand all
relationships.
Some different correlation Here’s how we calculate
calculations... correlation (Pearson’s way): Real life is never as simple as a coin toss.

There are different correlation calculations (called In this example we have two things to compare, X Probabilities ARE NOT Guarantees!
coefficients) for different kinds of data: and Y.
Probability tells you that over the long run there is
Pearson’s Coefficient – Measures linear 1. First calculate Mean (average) of X
a certain chance of something happening. Not that
relationship between two variables 2. Calculate Mean (average) of Y
something will or will not happen at a specific time.
3. Subtract Mean of X from each of X values (we’ll
Spearman’s Rank Coefficient – Measures
call these A), and subtract Mean of Y from each
relationship between two ordinal variables In other words, probabilities are great for general
of Y values (we’ll call these B)
Phi Coefficient – Measures relationship predictions about long term events, but they
4. Square A’s (we’ll call these C2’s)
between two dichotomous variables cannot and do not predict specific events.
5. Square B’s (we’ll call these D2’s)
6. Multiply all A’s by B’s (we’ll call these AB’s)
Pearson’s Coefficient is most popular and what 7. Add up all AB’s Over Generalization of Results
analytical tools use. 8. Add up all C2’s
9. Add up all D2’s If you calculate a correlation on a specific
10. Now perform calculation below… population, you cannot then say that correlation is
same for general population.
Sum of all AB’s
Correlation =
(Sum of C2’s) x (Sum of D2’s)

Make sure to review Hazards! section regarding correlation and causation! Check out our Statistics Cheat Sheet

Start Your Journey With Us:

(888) 252-7866 [email protected] | www.blastam.com
Locations: Rocklin, San Francisco, New York, Seattle, Los Angeles, Chicago, Boston, London