8:56 2014-10-07 Tuesday
random variables & their distribution
8:57 2014-10-07
Bernoulli(p) => Binomial(n,p)
8:58 2014-10-07
Binomial distribution
8:58 2014-10-07
X ~ Bin(n, p)
story: X is the #successes of n independent Bernoulli(p) trials
9:01 2014-10-07
indicator r.v.
9:02 2014-10-07
sum of random indicator variables:
X = X1 + X2 + ... + Xn
9:03 2014-10-07
i.i.d. == independent identically distributed
9:05 2014-10-07
confuse r.v. with distribution
9:06 2014-10-07
event is a subset of sample space
9:23 2014-10-07
event is a subset of outcomes
9:24 2014-10-07
CDF == Cumulative Distribution Function
9:25 2014-10-07
CDF is a way to describe the distribution.
9:46 2014-10-07
PMF is only for discrete r.v.
9:47 2014-10-07
continuous r.v.
9:48 2014-10-07
the reason we use CDF is that it's more general.
PMF is only for discrete r.v.s
9:53 2014-10-07
those are equally valid ways to describe the distribution.
9:53 2014-10-07
that's why it's called Binomial distribution,
because it's connected to the binomial theorem.
9:55 2014-10-07
X ~ Bin(n, p), Y ~ Bin(m, p) => X+Y ~ Bin(m+n, p)
9:59 2014-10-07
mathematically we're adding 2 functions
9:59 2014-10-07
P(X + Y = k) // in statistics, this is called convolution
10:06 2014-10-07
where we condition on X, and using the LOTP(Law Of Total Probability)
10:07 2014-10-07
independence mean we can just cross this, thus
P(A|B) = P(A)
10:10 2014-10-07
the key assumption is that the trials
are indepent & they all have the same probability of success.
10:15 2014-10-07
hypergeometric distribution
start Harvard statistics, video 8
random variables & their distribution
8:57 2014-10-07
Bernoulli(p) => Binomial(n,p)
8:58 2014-10-07
Binomial distribution
8:58 2014-10-07
X ~ Bin(n, p)
story: X is the #successes of n independent Bernoulli(p) trials
9:01 2014-10-07
indicator r.v.
9:02 2014-10-07
sum of random indicator variables:
X = X1 + X2 + ... + Xn
9:03 2014-10-07
i.i.d. == independent identically distributed
9:05 2014-10-07
confuse r.v. with distribution
9:06 2014-10-07
event is a subset of sample space
9:23 2014-10-07
event is a subset of outcomes
9:24 2014-10-07
CDF == Cumulative Distribution Function
9:25 2014-10-07
CDF is a way to describe the distribution.
9:46 2014-10-07
PMF is only for discrete r.v.
9:47 2014-10-07
continuous r.v.
9:48 2014-10-07
the reason we use CDF is that it's more general.
PMF is only for discrete r.v.s
9:53 2014-10-07
those are equally valid ways to describe the distribution.
9:53 2014-10-07
that's why it's called Binomial distribution,
because it's connected to the binomial theorem.
9:55 2014-10-07
X ~ Bin(n, p), Y ~ Bin(m, p) => X+Y ~ Bin(m+n, p)
9:59 2014-10-07
mathematically we're adding 2 functions
9:59 2014-10-07
P(X + Y = k) // in statistics, this is called convolution
10:06 2014-10-07
where we condition on X, and using the LOTP(Law Of Total Probability)
10:07 2014-10-07
independence mean we can just cross this, thus
P(A|B) = P(A)
10:10 2014-10-07
the key assumption is that the trials
are indepent & they all have the same probability of success.
10:15 2014-10-07
hypergeometric distribution