Stats Definitions - designed experiment (direct control -Measurement Error (eg bad qn/ 

Ogive/Cumulative % Polygon (x-

- Variable: characteristic of an item or over who gets treatment) Hawthorne effect) axis: variable of interest, y-axis:
individual - observational studies (no control) cumulative %)
- Data: set of individual values Organizing - 2 Numerical Variables
associated with a variable Types of Samples - Categorical  Scatter Plot
- Statistics: methods that help transform - Non-probability  Summary Table (1 variable)  Time Sequence
data into useful information  Judgement (get opinions of  Contingency Table (2 variables)
- Population: all the items/individuals experts) - Numerical Central Tendency
about which you want to draw a  Convenience (easy)  Ordered Array (rank from min to - Central Tendency: extent to which all
conclusion - Probability max) data values group around a central
- Sample: proportion of population  Simple Random (equal chance  Frequency Distribution value
selected for analysis of being picked) o Class, frequency, n N

- Population Parameter: summarizes  Systematic (pick every kth frequency %, cum -

∑ xi , ∑ xi
i=1
the value of a specific variable of a person, where frequency, cum % x= μ= i=1
population n N
number of people o 5-15 classes
- Sample Statistic: summarizes the value k= n+1
sample ¿ ¿ ¿ o - Median (position) =
of a specific variable for sample data 2
 Stratified (divide pop into strata range - Mode = most common value
class interval=
& select sample to mimics its no . of groups desired
Types of Variables - Should use both mean and median
characteristics)  Cumulative Distribution
- Categorical since mean is affected by extreme
 Cluster (divide pop into clusters;
 Nominal (defined categories)  outliers
each representative of pop) Visualizing
 nominal scale - Summary
 Ordinal (ordered categories)  Variation
Comparing Sampling Methods  Bar Chart
ordinal scale - Variation: amount of dispersion/
- Simple Random/Systematic  Pie/Doughnut Chart
- Numerical scattering of values
 Simple to use  Pareto Chart (bar chart with
 Discrete (counting) - Range = Xmax - Xmin (ignores distribution
 Not a good representative of decreasing order of frequency + of data & is sensitive to outliers)
 Continuous (measurement) population’s characteristics cumulative polygon) n N
 Uses either interval scale (no 0 - S =∑ ¿ ¿ ¿), σ =∑ ¿ ¿ ¿
2 2
- Stratified - Contingency
point) or ratio scale (true 0  Ensures representation  Side by Side Bar Chart i =1 i=1
n N
scale) - Cluster  Doughnut Chart - S=√ ∑ ¿ ¿ ¿ , σ =√ ∑ ¿ ¿ ¿
 Cost effective - Ordered Array i=1 i=1
Sources of Data  Less efficient S
 Stem-and-Leaf Display
- Data distributed by organizations or - Coefficient of Variation: CV = ×100
- Frequency/Cumulative Distribution x
individuals (eg weather report, financial Survey Errors  Histogram (x-axis: midpoint, y-  Always in %
statements) - Coverage Error/Selection Bias (exclude axis: freq/ rel freq/ freq %)  Measures relative variation to
- Survey people)  Polygon (x-axis: midpoint, y- mean
- Data collected by ongoing business - Nonresponse Error axis: freq %)
activities (eg Big Data) - Sampling Error (always exists)
  Can be used to compare Q2-min > Q2-min = Q2-min < - Empirical: based on data collected  No. of events in one area of
variability of 2/more data sets max-Q2 max-Q2 max-Q2 - Subjective: based on experience, opportunity is independent of
with different units Q1-min > Q1-min = Q1-min < opinion and analysis of a situation the no. of events in other areas
x −x max-Q3 max-Q3 max-Q3 of opportunity
- Z-score: Z=
S Q2-Q1 > Q3- Q2-Q1 = Q3- Q2-Q1 < Q3- Probability Formulas  Probability that 2 or more
 Extreme outlier if Z < -3.0 or Q2 Q2 Q2 - P(A U B) = P(A) + P(B) – P(A n B) events occur in an area of
Z > +3.0 *Just need to fulfill 2/3 conditions P ( A n B) opportunity approaches 0 as the
Measure of R/S Between 2 Variables - P(A|B) =
P (B) area of opportunity shrinks
Shape - Covariance: measures the strength of -If 2 events are independent, then P(A|  Avg no. of events = λ
- Shape: pattern of distribution the linear relationship X & Y B) = P(A)
−λ
e λ
x

- - P ( X=x|λ )=
- Skewness: measures extent to which Discrete Random Variables X!
n
data values are not symmetrical (X)
 Left-skewed (-ve):
∑ ( X i− X)(Y i −Y ) 1 - μ=E - μ= λ
- σ 2=λ
¿ - σ =E ( X ) −E(X )
2 2 2
mean<median<mode cov ( X , Y )= i=1 =
n−1 n−1 - σ =λ
 Symmetric:  cov(X,Y)>0  X & Y tend to Binomial Distribution
mean=median=mode move in the same direction - Characteristics: Normal Distribution
 Right-skewed (+ve): X−μ
 cov(X,Y)<0  X & Y tend to  Fixed number of observations - Z=
mode<median<mean σ
move in the opposite  Each observation is categorized
- Kurtosis: affects peakedness of curve - Empirical Rules:
direction into success or failure
 Constant probability 1. μ ± σ=0.6826
 cov(X,Y)=0  X & Y are
 Observations are independent 2. μ ± 2σ =0.9544
independent
- 3. μ ±3 σ=0.9973
- Coefficient of Correlation: measures
n! n −x - Evaluating Normality
P ( X=x|n , π )=
x
the relative strength of the linear π (1−π)
x ! ( n−x ) !  Construct charts/graphs (check
relationship between X & Y
- μ=E ( X )=nπ for symmetry and bell shape)
cov (X , Y )
- r=  Compute descriptive stats
S X SY - σ 2=nπ (1−π )
o Mean≈ Median≈ Mode
5 Number Summary *population coefficient of correlation = p - σ =√ nπ (1−π )
sample coefficient of correlation = r o IQR ≈ 1.33 σ
- min, Q1, Q2, Q3, max - Shape:
 Closer to -1  stronger -ve o Range ≈ 6 σ
n+1  π < 0.5  right-skewed
- Q1 (position) = relationship  Check theoretical properties
4  π > 0.5  left-skewed
 Closer to +1  stronger +ve o Empirical rule holds
n+1 o ~80% lies within
- Q2 (position) = relationship Poisson Distribution
2  Closer to 0  weaker linear x ± 1.28 σ
3(n+1) - Characteristics:
- Q3 (position) = relationship  Probability that an event occurs
4 Uniform Distribution
- IQR = Q3 – Q1 in one area of opportunity is the
Assessing Probability same for all areas of
Left- Symmetric Right- - Priori: based on prior knowledge opportunity
skewed skewed
- Uniform Distribution: probability
distribution that has equal probabilities
for all possible outcomes

- Probability = base x height

1
- f ( x )= if a ≤ X ≤ b
(b−a)
- Otherwise, f ( x )=0
a+ b
- μ=
2
2
(b−a)
- σ =√
12