Binomial Distribution- must satisfy bernouli trials
[Link] are only two possible outcomes for each trial
[Link] probability of success is the same for each trial
3. The outcome for different trials are independent
4. There are a fixed number n of Bernouli trials conducted
P(X=k) = b(k; n, p) = (nk) pk(1-p)n-k, k=0, 1, 2, 3, … , n
The number of objects k selected from a set of n objects
When quantity of trials is an inequality:
B(x; n, p) = k=0x b(k; n, p) for x = 0, 1, 2, 3, …, n
Normal Distribution
(x; , 2)= 1/((2)) e-(x-)^2 / 2^2
Standard Normal distribution
(z)=P(Z<=z)= 1/(2) -z e-t^2/2 dt
Standard Normal Probabilities
(Z) = P(Z<=z)
(b)-(a) = P(a < Z <= b)
Standardized random variable
Z= (X-)/
Expected Values
E(X) =
E(aX + b) = aE(X) + b
V(xbar) = 2/n
Sd(xbar) = /n
Central Limit Thrm
If xbar is the mean of a sample size n taken from a population having the mean and the
finite variance 2, then; Z = (Xbar - ) / (/n) is a rv whose distribution function
approaches that of the standard norm distributions as n approaches infinity
Confidence Intervals
Maximum error of estimate E = za/2 √(p(1-p)/n)
Sample Size determination n = p(1-p) [za/2/E]2
n = 1/4 [za/2/E]2
Confidence Intervals
Lg sample ci for [xbar – za/2 * /n < < xbar + za/2 * /n]
Or [xbar – za/2 * s/n < < xbar + za/2 * s/n]
Sm sample ci for [xbar – ta/2 * s/n < < xbar + ta/2 * s/n]
CI= xbar +- ta/2 * s/n
t = (xbar - )/(S/n)
Analysis of Variance
Analysis of Variance
Source of Degrees of Sum of Mean Squares F
variation freedom Squares
Treatments K-1 SS(Tr) MS(Tr)=SS(Tr)/(k-1) MS(Tr)/MSE
Error N–k SSE MSE=SSE/(N-k)
Total N–1 SST
SSE=SST-SS(Tr)
C= T.2/N
N= i=1k ni, quantity of sample sizes
T.= i=1k Ti and Ti= j=1ni yij, grand total of all sample sizes
SST= i=1k j=1ni y2ij – C, all samples squared minus C
SS(Tr)= i=1k(Ti2/ni) – C, samples added up in each treatment quared divided by sample
size minus C
Confidence Intervals
Lg sample confidence interval for the diff of 2 proportions
X1/n1 – x2/n2 +- za/2 [ (X1/n1*(1- X1/n1)/n1) + (X2/n2*(1- X2/n2)/n2)]
Confidence limits for a + bxo
(a + bxo) +- ta/2 * se * [(1/n) + (xo-xbar)2/ Sxx], number of degrees of freedom for ta/2 is n –
2
Probability Density Function
F(x) = P(X <= x) = -x f(t) dt
d/dx(F(x)) = f(x)
f(x)>= 0 for all x
- f(x) dx = 1
E(x) = - x * f(x) dx
Probability
P(AB) = P(A) + P(B) – P(AB)
Confidence Intervals mean differences
Xbar – ybar +- za/2 [(s12/n1) + (s22/n2)]
Hypothesis
1 - 2 = , norm pop, 1 and 2 not known, n1, n2 >= 30
Alternative Hypothesis Reject null hypothesis if
1 - 2 < , Z < -za
1 - 2 > , Z > za
1 - 2 <> , Z < -za/2 or Z > za/2
1 - 2 = , norm pop, 1 = 2, two sample t test
Alternative Hypothesis Reject null hypothesis if
1 - 2 < , t < -ta
1 - 2 > , t > ta
1 - 2 <> , t < -ta/2 or t > ta/2
= 0, lg sample
Alternative Hypothesis Reject null hypothesis if
< 0 Z < -za
> 0 Z > za
<> 0 Z < -za/2 or Z > za/2
= 0 norm pop, unknown, one sample t test
Alternative Hypothesis Reject null hypothesis if
< 0 t < -ta
> 0 t > ta
<> 0 t < -ta/2 or t > ta/2
Poisson Distribution- model for counts that do not have an upper bound
= np
f(x; ) = xe-/ x!, >0
x=0 f(x; )