AP Statistics Review Notecards

DESCRIPTIVE STATISTICS
Categorical Variable: data based on a name or label (e.g. color).
Quantitative Variables: numerical data, a measurable value (e.g. height).
Categorical Variables
Quantitative variables
Bar Graph
Dotplot
*
*
*
*
*
*
*
*
Pie Chart
*
Stems
Red
150
140
130
120
110
100
90
80
Stem plot
*
*
*
*
*
Leaves
Yellow
1
Orange
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
Green
Blue
Indigo
Violet
26
4579
12225799
0234457899
11478
Key: 110 7 represents an IQ score of 117
Box plot
Histogram
Patterns of Distribution:
Resistant: not affected by outliers.
Percentile: the pth percentile of a distribution is the value such that p percent of
observations fall below it.
Mode: most occurring value, signified by a peak on a graph.
ο‚· Shape
-Symetric
Unimodal
Bimodal
-Uniform
-Skewed
Left
Right
Center
-Median (50th Percentile): middle of the observations (resistant).
1
-Mean: 𝑛 βˆ‘ π‘₯ (non-resistant).
ο‚· Spread
-Range: difference between minimum and maximum value (non-resistant). It is
a single number, not an interval.
-Inner Quartile Range(IQR): Q3 – Q1 (resistant).
-Variance (Οƒ2-population/s2-sample): average of squared distances from the
1
1
mean (non-resistant). 𝜎 2 = 𝑛 βˆ‘(π‘₯ βˆ’ πœ‡)2 and 𝑠 2 = π‘›βˆ’1 βˆ‘(π‘₯ βˆ’ π‘₯Μ… )2 .
-Standard Deviation (Οƒ-population/s-sample): square root of variance (non
resistant).
ο‚· Outliers
-Outlier on the lower end: π‘₯ < 𝑄1 βˆ’ 1.5𝐼𝑄𝑅
-Outlier on the upper end: π‘₯ > 𝑄3 βˆ’ 1.5𝐼𝑄𝑅
Z-score: standardized value of x. 𝑧 =
Normal Distribution
ο‚·
π‘₯βˆ’πœ‡
𝜎
Empirical Rule:
𝑃(|𝑧| < 𝜎)=0.68
𝑃(|𝑧| < 2𝜎)=0.95
𝑃(|𝑧| < 3𝜎)=0.997
Linear Regression
Correlation Coefficient (r): standardized measure of the strength of association of bivariate quantitative linear data. π‘Ÿ =
𝑦𝑖 βˆ’π‘¦
Μ…
1
π‘₯𝑖 βˆ’π‘₯Μ…
βˆ‘( )( )
π‘›βˆ’1
𝑠π‘₯
𝑠π‘₯
βˆ’1 ≀ π‘Ÿ ≀ 1 (-1=perfect negative correlation; 0=absolutely no correlation; 1=perfect positive correlation)
Correlation DOES NOT imply causation!
Coefficient of Determination(R2): the fraction of variability of y-variable that is explained by the linear relationship of yvariable on x-variable.
Least-Squares Regression Line: line that minimizes the sum of the squares of the distances of all the points from the
line. LSRL passes through the point (π‘₯,
Μ… 𝑦̅).
𝑦̂ = π‘Ž + 𝑏π‘₯
𝑠𝑦
𝑏=π‘Ÿ
𝑠π‘₯
π‘Ž = 𝑦̅ βˆ’ 𝑏π‘₯Μ…
Extrapolation: prediction of values outside of the domain of LSRL. It is often not accurate.
Residual: the difference between the observed value and the predicted value. 𝑒̂𝑖 = 𝑦𝑖 βˆ’ 𝑦̂𝑖 , βˆ‘π‘›π‘–=1 𝑒̂𝑖 = 0, and 𝑠𝑒 2 =
1
βˆ‘ 𝑒2.
π‘›βˆ’2
When evaluating LSRL, look at the scatter plot, the residual plot, and R2. High R2 doesn’t imply a linear relationship. If
the residual is not randomly scattered (i.e. there is a clear pattern), reject the linear model.
Cautions about Correlation:
Lurking Variable: a variable that influences the interpretation of a model.
Direct Causation: change in x influences change in y.
Common Response: change in x and y cause by another variable.
Confounding: the effect of x is confounded by a lurking variable.
Experimental Design
Observational Study: observes individuals and measures variables of interest without
imposing a treatment(e.g. sample surveys).
Experiment: imposes a treatment on individuals to observe their responses.
Factor: the explanatory variable. It has at least two levels (specific values).
Experimental Unit: an individual on which the experiment is done (human experimental
units are subjects).
Treatment: specific experimental condtion applied to to a unit.
Control Group: a group that the other factors are compared against (old treatment or
placebo).
Avoiding Bias:
Blind or Double-Blind Experiement
Principles of Experimental Design:
ο‚· Control the effects of the lurking variable by comparing treatments to the control
group.
ο‚· Randomize the assignment to treatments.
ο‚· Replicate each treatment on many units to remove natural variation.
The ONLY way to show causation is with a well-designed, well-controlled experiment.
Experimental Designs:
Matched Pairs: each unit receives both treatments in a random order.
Blocking: units are assigned to blocks based on similar characteristics and then randomly
assigned to treatments.
Sampling
Population: an entire group of individuals.
Census: attempts to obtain information about every individual in the population of
interest.
Sample: a part of the population of interest.
Sampling: studying a part of a population in order to gain information about the whole.
Strata: group of similar individuals.
Sampling Methods:
Voluntary
-consists of individuals who chose themselves to respond.
Response
β†’biased (people with negative opinions are more likely to
respond).
Convenience
-a sample that chooses inidviduals that are easiest to reach
Sampling
β†’biased based on time and place
Probability
-sample chosen by chance
Sample
-(SRS, Systematic Random Sample)
Simple Random
-of size n consists of n individuals from a population chosen in
Sample (SRS)
such a way that every set of n individuals.
Stratified
-divides population into strata, an SRS is taken in each stratum,
Random Sample
and then the SRSs are combined to form a full sample.
Multistage
-combines several sampling techniques in the selection of the
Sampling
sample.
Cluster Sample
-individuals are chosen in groups that are similar in location.
Bias: systematic favoring of an outcome.
Response bias: caused by the behavior of the respondent or the interviewer (lying,
question wording, race or sex of the interviewer).
Undercoverage: group(s) from the population being left out from the sample-selection
process.
Nonresponse: an individual chosen for the sample cannot be contacted or does not
cooperate.
Probability
Random phenomenon: individual outcomes are uncertain though there is a regular
distribution of outcomes in a large number of repetitions.
Probability: measure of likelihood of an event.
Event: set of outcomes of a random phenomenon.
Sample Space S: a set of all possible outcomes of a random phenomenon.
Probability Rules:
1. 0 ≀ 𝑃(𝐴) ≀ 1
2. 𝑃(𝑆) = 1
3. Complement (Not A): 𝑃(𝐴𝐢 ) = 1 βˆ’ 𝑃(𝐴)
4. Intersection (A and B): 𝑃(𝐴 ∩ 𝐡) = 𝑃(𝐴)𝑃(𝐡|𝐴)
5. Union (A or B): 𝑃(𝐴 βˆͺ 𝐡) = 𝑃(𝐴) + 𝑃(𝐡) βˆ’ 𝑃(𝐴 ∩ 𝐡)
P(A∩B)
6. Conditional Probability (A Given B): P(A|B) =
P(B)
7. Independence: 𝑃(𝐴|𝐡) = 𝑃(𝐴) β‡’ 𝑃(𝐴 ∩ 𝐡) = 𝑃(𝐴)𝑃(𝐡)
8. Disjoint: 𝑃(𝐴 ∩ 𝐡) = 0 β‡’ 𝑃(𝐴 βˆͺ 𝐡) = 𝑃(𝐴) + 𝑃(𝐡)
Sampling with Replacement: P(A) is constant for all successive trials.
Sampling without Replacement: P(A) changes for successive trials.
Random Variables
Random Variable X: variable whose value is based on the outcome of a random event.
Discrete Random Variable: random variable with a finite number of possible values. 𝑃(𝑋 =
π‘₯) = 𝑝
Continuous Random Variable: random variable whose sample space is an entire
interval. 𝑃(𝑋 = π‘₯) = 0 ; 𝑃(π‘₯1 ≀ 𝑋 ≀ π‘₯2 ) = 𝑝.
Normal Distribution N(ΞΌ,Οƒ): normal distribution with mean ΞΌ and standard deviation Οƒ.
Expected Value (𝝁𝑿 ): mean of random variable. βˆ‘ π‘₯𝑖 𝑝𝑖
Variance of Random Variable (πˆπ‘Ώ 𝟐 ): βˆ‘(π‘₯𝑖 βˆ’ πœ‡π‘₯ )2 𝑝𝑖
Standard Deviation of Random Variable (πˆπ‘Ώ ): βˆšβˆ‘(π‘₯𝑖 βˆ’ πœ‡π‘₯ )2 𝑝𝑖
Law of Large Numbers: the greater the sample size, the closer π‘₯Μ… will get to πœ‡π‘‹ .
Mean Rules
Variance Rules
πœ‡π‘‹+π‘Œ = πœ‡π‘‹ + πœ‡π‘Œ
𝜎 2𝑋±π‘Œ = 𝜎 2𝑋 + 𝜎 2 π‘Œ
πœ‡π‘‹βˆ’π‘Œ = πœ‡π‘‹ βˆ’ πœ‡π‘Œ
𝜎 2 π‘Ž+𝑏𝑋 = 𝑏 2 𝜎 2𝑋
Standard Deviation Rule
πœ‡π‘Ž+𝑏𝑋 = π‘Ž + π‘πœ‡π‘‹
πœŽπ‘‹±π‘Œ = √𝜎 2𝑋 + 𝜎 2 π‘Œ
Binomial Model
Bernoulli Process:
ο‚·
A finite sequence of Bernoulli Events.
ο‚·
Each event has two possible outcomes.
ο‚·
Each event is independent.
ο‚·
Each event has the same probability.
Binomial Distribution B(n,p): distribution of the count X of successes with probability p in the binomial setting
with n observations.
Complement of p (q): the probability of a failure. 1-p
𝒏!
Permutation π’π‘·π’Œ : the number of different arrangements where order does matter.
π’Œ!
Combination 𝒏π‘ͺπ’Œ : the number of different arrangements where the order does not matter. (π’π’Œ) =
𝒏!
π’Œ!(π’βˆ’π’Œ)!
Binomial Probability: 𝑃(𝑋 = π‘˜) = (π‘›π‘˜)π‘π‘˜ (π‘ž)π‘›βˆ’π‘˜
Normal Approximation of Binomial Distribution: Suppose B(n,p), then when npβ‰₯10 and nqβ‰₯10, the
distribution of X can be approximated by the normal model N(𝑛𝑝, βˆšπ‘›π‘π‘ž).
(𝑿 = 𝒏)
𝑷(𝑿 > 𝒏)
𝝁𝑿
𝝈𝟐 𝑿
Geometric Model
Probability of X occurring on nth trial
Probability of X occurring after nth trial
Mean of a Geometric Count
Variance of a Geometric Count
= π‘π‘žπ‘›βˆ’1
= π‘žπ‘›
= π‘βˆ’1
π‘ž
= 2
πˆπ‘Ώ
Standard Deviation of a Geometric Count
=√
𝑝
π‘ž
𝑝
Sampling Distributions
Parameter: describes a population
Statistic: describes a sample
Sampling distribution: distribution of values taken by a statistic in all possible samples of the same size from the same
population
Unbiased statistic: the mean of a sampling distribution is equal to the true parameter being estimated
Variability of a statistic: spread of a sampling distribution
Notation
p
Μ‚
𝒑
Population Proportion
Sample Proportion
𝝁𝒙̅
πˆπ’™Μ… 𝟐
Mean of Sampling Distribution
Variance of Sampling Distribution
πˆπ’™Μ…
Standard Deviation of Sampling Distribution
=
𝝁𝒑̂
πˆπ’‘Μ‚ 𝟐
Mean of Sampling Distribution of Sample Proportion
Variance of Sampling Distribution of Sample Proportion
=𝑝
π‘π‘ž
=
πˆπ’‘Μ‚
Standard Deviation of Sampling Distribution of Sample Proportion
𝑋
=
𝑛
=πœ‡
=
𝜎2
𝑛
𝜎
βˆšπ‘›
𝑛
=√
π‘π‘ž
𝑛
{
π‘π‘œπ‘π‘’π‘™π‘Žπ‘‘π‘–π‘œπ‘› 𝑠𝑖𝑧𝑒 β‰₯ 10𝑛
𝑛𝑝 β‰₯ 10
π‘›π‘ž β‰₯ 10
Properties of Sampling Distributions:
ο‚·
Sampling distribution of π‘₯Μ… is normal if the population distribution is normal.
ο‚·
Central Limit Theorem: The sampling distribution of π‘₯Μ… is close to the normal distribution N (ΞΌ,
ο‚·
n is sufficiently large (𝑛 β‰₯ 30), even when the population distribution is not normal.
Variability decreases as sample size increases.
Inference
Significance Test:
Null Hypothesis (π‡πŸŽ ): claim about a parameter that is initially believed to be true.
Alternate Hypothesis (𝐇𝒂 ): claim competing with the null hypothesis.
P-Value: the probability of obtaining the observed test statistic assuming the null
hypothesis is true.
Significance Level (Ξ±): the probability threshold beyond which we reject the null
hypothesis
Effect: the difference between true and hypothesized mean.
Power : Increasing sample size, effect size, or significance level will increase
power.
Truth about Population
H0 True
Hπ‘Ž True
Type I Error
Correct Decision
Reject H0
Decision
(P=Ξ±)
(P=1-Ξ²)
Correct Decision
Type II Error
Accept H0
(P=1-Ξ±)
(P=Ξ²)
Z Statistic
Conditions
INDEPENDENCE
RANDOMIZATION: SRS
TEN PERCENT: π‘π‘œπ‘π‘’π‘™π‘Žπ‘‘π‘–π‘œπ‘› 𝑠𝑖𝑧𝑒 β‰₯ 10𝑛
NORMALITY: 𝑛𝑝 β‰₯ 10 and π‘›π‘ž β‰₯ 10
Confidence Interval
𝐳 βˆ— : confidence level.
One Proportion Z Interval
𝑝̂ ± 𝑧 βˆ— βˆšπ‘Μ‚ π‘žΜ‚β„π‘›
Two Proportion Z Interval
𝑝̂1 π‘žΜ‚1 𝑝̂2 π‘žΜ‚2
𝑝̂1 βˆ’ 𝑝̂2 ± 𝑧 βˆ— √
+
𝑛1
𝑛2
Significance Tests
One Proportion Z Test
𝑧=
π‘―πŸŽ : 𝑝 = 𝑝0
𝑯𝒂 : 𝑝 > 𝑝0
𝑯𝒂 : 𝑝 < 𝑝0
𝑯𝒂 : 𝑝 β‰  𝑝0
Two Proportion Z Test
𝑃(𝑍 > 𝑧)
𝑃(𝑍 < 𝑧)
2𝑃(𝑍 > |𝑧|)
𝑝̂1 βˆ’ 𝑝̂2
𝑧=
√
π‘―πŸŽ : 𝑝1
𝑯𝒂 : 𝑝1
𝑯𝒂 : 𝑝1
𝑯𝒂 : 𝑝1
= 𝑝2
> 𝑝2
< 𝑝2
β‰  𝑝2
𝑝̂ βˆ’ 𝑝0
βˆšπ‘0 π‘ž0 ⁄𝑛
𝑝̂1 𝑛1 + 𝑝̂2 𝑛2
𝑛1 + 𝑛2
𝑃(𝑍 > 𝑧)
𝑃(𝑍 < 𝑧)
2𝑃(𝑍 < |𝑧|)
𝜎
βˆšπ‘›
) when
Inference
Confidence Interval:
Estimate ± Margin of Error
Confidence Level: the probability that the method will give a correct
answer.
Margin of Error: decreasing the confidence level, deacrasing the standard
deviation, and increasing the sample size will decrease the margin of error.
Significance Test Procedure:
1.
Name the test and state 𝐻0 and π»π‘Ž .
2.
Check conditions.
3.
Perform appropriate calculations.
P > Ξ± β†’ Accept 𝐻0
P ≀ Ξ± β†’ Reject 𝐻0
4.
Conclusion.
T Statistic
Conditions
INDEPENDENCE
RANDOMIZATION: SRS
NORMALITY or 𝑛 β‰₯ 40
Confidence Interval
𝐭 βˆ— : confidence level.
One Sample/Matched Pairs T Interval
𝑠
π‘₯Μ… ± 𝑑 βˆ—
βˆšπ‘›
Two Sample T Interval
π‘₯Μ…1 βˆ’ π‘₯Μ…2 ± 𝑑 βˆ— √
𝑠1 2 𝑠2 2
+
𝑛1
𝑛2
Significance Tests
One Proportion/Matched Pairs T Test
π‘₯Μ… βˆ’ πœ‡0
𝑑=
𝑠
βˆšπ‘›
π‘―πŸŽ : πœ‡ = πœ‡0
𝑯𝒂 : πœ‡ > πœ‡0
𝑃(𝑇 > 𝑑)
𝑯𝒂 : πœ‡ < πœ‡0
𝑃(𝑇 < 𝑑)
𝑯𝒂 : πœ‡ β‰  πœ‡0
2𝑃(𝑇 > |𝑑|)
Two Sample T Test
π‘₯Μ…1 βˆ’ π‘₯Μ…2
𝑧=
𝑠 2 𝑠 2
√ 1 + 2
𝑛1
𝑛2
π‘―πŸŽ : πœ‡1 = πœ‡2
𝑯𝒂 : πœ‡1 > πœ‡2
𝑯𝒂 : πœ‡1 < πœ‡2
𝑯𝒂 : πœ‡1 β‰  πœ‡2
𝑃(𝑇 > 𝑑)
𝑃(𝑇 < 𝑑)
2𝑃(𝑇 > |𝑑|)