Formulas That May Be Needed

1 Laws of Probability
• If A and B are mutually exclusive events, then P (A or B) = P (A) + P (B).

• If A and B are independent events, then

– P (A and B) = P (A) × P (B),

– P (A | B) = P (A).

• If P (A and B) = P (A) × P (B) or P (A | B) = P (A) or P (B | A) = P (B), then

– A and B are independent events.

• If A and B are two events and P (B) =

√ 0, the conditional probability of A given B is

P (A and B) P (B | A) × P (A)
P (A | B) = = .
P (B) P (B)

2 Discrete Random Variables (RV from now on)

n n
2
X̄ = E(X) = µX = P (X = xi ) xi VAR(X) = �X = P (X = xi )(xi − µX )2
i=1 i=1
�
Std Dev(X) = �X = VAR(X)

3 Binomial Distribution with Parameters n and p

n!
µX = np 2
�X = np(1 − p) P (X = x) = px (1 − p)n−x , x = 0, 1, . . . , n
x!(n − x)!

4 Two Random Variables

n
Cov(X, Y ) = P (X = xi and Y = yi )(xi − µX )(yi − µY )
i=1
Cov(X, Y )
Corr(X, Y ) =
�X �Y

If X and Y are independent, then Cov(X, Y ) = 0 and Corr(X, Y ) = 0.

E(aX + bY + c) = aE(X) + bE(Y ) + c

VAR(aX + bY + c) = a2 VAR(X) + b2 VAR(Y ) + 2ab Cov(X, Y )

= a 2 �X
2
+ b2 �Y2 + 2ab �X �Y Corr(X, Y )
5 Uniform Distribution between a and b
a+b (b − a)2 x−a
E(X) = VAR(X ) = P (X � x) = if a � x � b
2 12 b−a

6 Normal Distribution
• If X is a normal distribution with mean µ and standard deviation �, then P (X � x) =
F ( x−µ x−µ
� ), where F (z) can be read from the “normal” table and z = � .

• If X and Y are Normally distributed, then so is aX + bY + c.

• Assume that X� 1 , .., Xn are independent and identically distributed, E(Xi ) = µ, and VAR(Xi ) =
� 2 . Let Sn = ni=1 Xi be the sum and X ¯ = 1 �n Xi be the average, then:
n i=1

– Central Limit Theorem for the sum. If n is moderately large (say, 30 or more) then
�
Sn is approximately Normally distributed with mean nµ and standard deviation � n.
– Central Limit Theorem for the sample mean. If n is moderately large, then X ¯ is
�
approximately Normally distributed with mean µ and standard deviation �n .

• A binomial distribution can be approximated with a normal (with the correct parameters µ
and �) when np � 5 and n(1 − p) � 5.

7 Statistical Inference for the Population Mean µ

• X¯ = 1 �n Xi is the sample mean. The observed sample mean x ¯ = n1 ni=1 xi is an estimate
�
n i=1
of the mean of the population µ.
�
• S = n−1 1 �n ¯ 2
i=1 (Xi − X) is the standard deviation of the sample. The observed standard
�
1 �n 2
deviation of the sample s = n−1 i=1 (xi − x̄) is an estimate of the standard deviation of
the population �.

• The standard deviation of the sample mean is Std Dev(X ¯ ) = �� , where � is the standard
n
deviation of the population.
¯ −µ
X
• If n is large (say, 30 or more), then is approximately a standard Normal RV.
�S
n

¯ −µ
X
• If n is small (say, less than 30) and the population distribution is “well-behaved”, then
�S
n
obeys a t-distribution with n − 1 degrees of freedom (dof).
� �
s s
¯ − c × � ,x
• For n � 30 an �% confidence interval for the real mean µ is x ¯+c× � ,
n n
where c can be found by solving P (−c � Z � c) = �/100 with Z being a standard Normal
RV. For example:

For � = .90, c = 1.645; for � = .95, c = 1.960; for � = .98, c = 2.326; for � = .99, c = 2.576.
 • For n < 30 and�a “well-behaved” population
� distribution, an �% confidence interval for the
s s
¯ − c × � ,x
real mean µ is x ¯+c× � , where c satisfies that P (−c � T � c) = �/100
n n
with T a RV that has a t-distribution with n − 1 dof.

• To construct an �% confidence interval that is within (plus or minus) L of the actual mean,
c2 s 2
the required sample size is n = 2 , where c satisfies that P (−c � Z � c) = �/100 if Z is a
L
standard Normal RV.

8 Regression
• n = number of data points

• k = number of explanatory (independent) variables

• Based on observed data

(y1 , x11 , . . . , xk1 )
...
(yn , x1n , . . . , xkn )

• Population relation: Yi = �0 + �1 x1i + . . . + �k xki + αi where αi is N (0, �)

• ŷi = b0 + b1 x1i + . . . + bk xki is the predicted value

• bj is the regression coefficient and an estimate of �j , j = 0, 1, . . . , k

• sbj is the standard deviation of bj

• ei = yi − ŷi is the residual

• An �% confidence interval for �j is [bj − c × sbj , bj + c × sbj ]

where c satisfies that P (−c � T � c) = �/100 if T obeys a t-distribution with dof = n − k − 1
bj
• The t-statistic is tbj =
s bj

• Checklist for evaluating a linear regression model: (i) linearity, (ii) signs of regression co­
efficients, (iii) significance of independent variables, (iv) R2 , (v) normality of the residuals,
(vi) heteroscedasticity, (vii) autocorrelation, and (viii) multicolinearity.