AP Statistics 2023 Cheat Sheet | See all AP Stats study guides | @thinkfiveable

Unit 1 Unit 2 Unit 3 Unit 4

Probability, Random Variables, & Probability
Exploring One-Variable Data Exploring Two-Variable Data Collecting Data
Distributions

Probability = the chance of an event

occurring, expressed in a decimal 0 1
P(event) = successful outcomes / total
outcomes
Categorical data (not numerical) is
Complement of an event P(not event) is
shown in two-way tables & bar For categorical data to be independent, Simple Random Sample SRS =every group of
equal to 1 P(event)
graphs, analyzing proportions conditional a certain size has an equal chance of being
Common complements = at
Quantitative data is displayed in frequency = marginal frequency selected
least, at most, greater/less
histograms, dotplots, box plots, stem For quantitative data, always describe Cluster Sample Divide pop. into
than
and leaf plots, and scatterplots. associations with heterogeneous groups [all from some]
P A and B P A B probability that
Mean: non-resistant (affected by direction, strength, form. Stratified Random Sample Divide pop. into
BOTH events A and B occur
outliers) DIrection - positive / negative (slope) strata of homogeneous groups [some from all]
In a 2-way table = intersection
Median: resistant (affected by Form - linear / non-linear Why stratify by X, explain why indivs in those
of the two events divided by
outliers) r (correlation coefficient) measures strength & strata would have different rsps as opposed to
full total of the two events
Unimodal = one clear peak, Bimodal = direction, NOT FORM some other variable
Using condt’l probability P A
two clear peaks, Uniform = no clear Least Squares regression line LSRL predicts Stratifying: variability, precision
* P B|A
peaks, flat values of response variable (y) given Bias types=undercoverage, nonresponse,
P A or B PAUB probability that
Use comparison words when explanatory variable (x) response bias (inaccurate)-Always say if it
either events A or B occur
comparing distributions LSRL written as = a + bx leads to over/underestimate of a rsp
PA P B P A and B
For a histogram -> make sure you = predicted value of rsp variable EXPERIMENTS ASSIGN TREATMENTS
Conditional Probability P A|B PA
approximate the mean 500 750 a = y-int, b = slope Confounding-When a variable and the exp.
given B probability that event A
units) and use words like “no more” / Residual = predicted - actual variables are associated in a way that their
occurs given B already happened
“approximately” when describing Look for random scatter on residual plot! effects on a rsp. Variable can’t be
P A|B P A and B / P B
range Using LSRL to make predictions outside the distinguished from one another
Events are mutually exclusive if P A or B
When analyzing distributions, always interval of values of x used to make the Experiments have comparison, random
PA PB
CUSS & BS in context Center, equation of the line=extrapolation assignment (creates roughly equiv. groups of
Events are indep if P A|B P A OR if
Unusual features, Shape, Spread S & R-sq influenced by outliers (s , r-sq ) exp. units by balancing the effects of other
P A and B P A *P B
(remember skew pulls mean), BE Key Interpretations: For FRQ Writing) variables among treatment groups), control
Random variables are quantitative and
SPECIFIC and always contextualize Slope/b As the [exp var.] increases by 1 [unit], (helps avoid confounding & variability in rsp
take numerical values determined by the
your answer. the [rsp var.] is predicted to increase by b var.), & replication (any diffs in effects of
outcome of a chance event.
Normal distribution: mound-shaped [units]. treatments can be distinguished from chance
Discrete = has a set number of values
and symmetric. Its parameters are mu Y-intercept When there are zero [exp var], the differences b/w groups)
like number of coins that land heads or
( ) for mean and sigma for standard predicted [rsp var.] is y-int. Randomized block design: random assignment
composite ACT score
deviation. s When using LSRL to predict [rsp var] from of treatments is carried out separately in each
P X event) must add up to 1
Calculate z-score (value-mean / SD , [exp var] we are typically off by [value of s]. block
Expected Value OR MEAN of discrete
measuring how many SD a value is r² (in % About [r-sq]% of variation in [rsp var] Blocks share a var that may impact rsp
random variable is calculated using
from a mean is explained by the LSRL using [exp var]. variability in rsp var, allows for easier
(x_1* p_1 (x_2*p_2 ..+(x_i*p_i)
The Standard Normal Distribution has a residual The actual (rsp var) is about comparison of treatments
Continuous random var takes on all
a Mean of 0 and a SD of 1 [residual] more/less than the predicted (rsp Matched pairs = compare 2 treatments in
values in an interval of numbers,
Empirical Rule: 68% of observations var). block size 2
represented by normal distribution
within 1 S
Binomial Random Variables: multiple
trials of the same event
Conditions: Binary,
Independent 10% , Number of

FRQ Tips: Allocate your time wisely for answering open-ended questions / Keep in mind that your answers will be evaluated as a whole / Familiarize yourself with
statistical terminology and use it accurately / Highlight the keywords in the questions by underlining or circling them / Identify and verify all underlying assumptions /
Be proficient in creating graphs manually and analyzing data presented in different forms.