Review of Probability, Random Variables, and Distributions
Lecture 8
Definition of Variance
- Def: The variance of
is: If
is discrete If
is continuous is called the standard deviation
- Properties
If
and are independent, then If
are mutually independent,
Covariance
Def: the covariance of
and is If
and are independent, The inverse direction may not be true!
Correlation coefficient
Def: The correlation coefficient of
and is If
and are independent, Properties:
pf: use
If
, we call and are completely linear correlated If
, and are called uncorrelated, means there is not any "linear correlation" between and . independent
uncorrelated
denote the strongness of linear correlation between and means there is positive linear correlation between and If
becomes larger, then tends to become stronger
Lecture 9
Bernoulli Distribution
0-1 distribution
Indicator
It can be used everywhere
Binomial Distribution
- Def: the number of success
in Bernoulli trails - If
, it becomes Bernoulli distribution : - Binomial:
- hint:
are mutually independent
- hint:
Multinomial Distribution
- Def: Multinomial experiments repeatedly: independent,
outcomes each time - DefL Multinomial distribution: the number of each outcomes in
trails - Joint
: - Each marginal distribution is binomial
Lecture 10
Hypergeometric Distribution
- Motivation: Sampling without replacement
- Def:
the number of success is selected from terms without replacement; - of
terms, are success and are failures.
:
Relationship to Binomial
- Binomial is the limit case for hypergeometric when
approaches infinity - When
is larger enough( is small):
- Binomial is the limit case for hypergeometric when
is hypergeometric with , then
Multivariate Hypergeometric
N terms be Lectureified into k kinds, select n randomly, number of each kind
Each marginal is hypergeometric!
Geometric Distribution
Def: Do Bernoulli experiments until succeed,
the number of trails pmf:
Mean
and variance
Negative Binomial Distribution
Def: Do Bernoulli experiments until the k-th succeed,
the number of trails pmf:
Mean
and variance
Poisson Distribution
Def: number of occurring in a Poisson process
Derivation: Poisson theorem
pmf:
Expectation:
Relationship to Binomial
- Poisson distribution is the limit case of binomial when
approaches infinity while is fixed - If
is large while is small,
- Poisson distribution is the limit case of binomial when
Lecture 11
Uniform Distribution
Def:
is called uniform distribution on if its density satisfy: cdf and probability
Expectations:
Exponential Distribution
Def:
is called exponential distribution if cdf:
Gamma Distribution
Gamma Function
Def: Gamma function
Properties:
Def: the Gamma density is as following:
Exponential is special case of Gamma density
Expectations:
Normal Distribution
Standard Normal
Def:
is called standard normal if density The cdf can be found from tables
Expectations: if
is standard normal Def:
is normal with parameter The density of
is: Expectations:
pth quantile
- Def: for p in
, the pth quantile of is - Def: for p in
, the critical value of is
- Def: for p in
Lecture 12
Central Limit Theorem
Th (Lindeberge-Levy): if
is a iid sequence with Then
Lecture 13
Estimation Methods
Moment estimate
Fundamental basis:
iid Distribution parameter
is related to Estimation:
The Method of Maximum Likelihood
Suppose the population
is called likelihood function The estimation of mle is chosen as:
Solution of mle for uniform distribution
find the likelihood function for
find mle
The likelihood function is strictly increasing with
but strictly decreasing with , so the mle are:
Lecture 14
Unbiasedness
- Def: if
, is called unbiased - Def:
is called bias - Def: if
, is asymptotically
Efficiency
- Def: both
and are biased, is more efficient than if
Mean Squared Error(MSE)
Def: the mean squared error is:
The MSE can be computed as:
Lecture 15
Chi-Squared Distribution
Derive of density:
Expectations:
Chi-Squared distributions are addictive:
t-Distribution
Density:
Even function
Limit is standard normal:
F-Distribution
- Property:
- The limit case is Normal Distribution
Sampling Distribution Theorems
Suppose the population is Normal:
Th1:
Th2:
and are independent, and Th3:
Lecture 16
CI under Normal Distribution
- find
, and is given - find
- construct
- find
- solve
- find
, and is unknown - find
- construct
- find
- solve
- find
- find
, and is given - construct
- solve
- construct
, and is unknown - construct
- construct
Sampling Distribution under Two Populations
Suppose
, independent, samples from Th1: var known
Th2: var unknown but equal
- Th3: Sampling theorem for Variance