Scribd
Upload
217 views4 pages

Statistic Formulas

This document summarizes common statistical formulas used to calculate variance, covariance, estimators, and distributions. It provides formulas for population and sample parameters like the population mean (μ), variance (σ2), and standard deviation. It also lists the distributions that common estimators follow, such as the sample proportion (p^) following a normal distribution. Confidence intervals and hypothesis tests are described using pivotal quantities, t-distributions, chi-square distributions, and the F-distribution.

Uploaded by

Basoko_Leaks
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
217 views4 pages

Statistic Formulas

This document summarizes common statistical formulas used to calculate variance, covariance, estimators, and distributions. It provides formulas for population and sample parameters like the population mean (μ), variance (σ2), and standard deviation. It also lists the distributions that common estimators follow, such as the sample proportion (p^) following a normal distribution. Confidence intervals and hypothesis tests are described using pivotal quantities, t-distributions, chi-square distributions, and the F-distribution.

Uploaded by

Basoko_Leaks
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
You are on page 1/ 4

STATISTIC FORMULAS

∑𝑥𝑥𝑖𝑖2 − 𝑛𝑛(𝑥𝑥̅ )2 Covariance ∑𝑥𝑥𝑖𝑖 𝑦𝑦𝑖𝑖 − 𝑛𝑛(𝑥𝑥̅ )(𝑦𝑦�)


Variance [𝝈𝝈𝟐𝟐 ]
𝑛𝑛 − 1 [𝒄𝒄𝒄𝒄𝒄𝒄(𝒙𝒙, 𝒚𝒚)] 𝑛𝑛 − 1

POPULATION � 𝑿𝑿 � = 𝑬𝑬�𝜽𝜽
� 𝑿𝑿 � − 𝜽𝜽𝑿𝑿
ESTIMATOR Bias�𝜽𝜽 UNBIASED ESTIMATOR?
PARAMETER

Population mean [𝝁𝝁] 𝑥𝑥̅ 𝐸𝐸[𝑥𝑥�] − 𝜇𝜇𝑋𝑋 = 0 Yes, if X normal

Population prop. [𝒑𝒑] 𝑝𝑝̂𝑋𝑋 𝐸𝐸[𝑝𝑝�] − 𝑝𝑝𝑋𝑋 = 0 Yes

Pop. variance [𝝈𝝈𝟐𝟐 ] 𝜎𝜎�𝑋𝑋2 𝐸𝐸�𝜎𝜎�2𝑋𝑋 � − 𝜎𝜎2𝑋𝑋 ≠ 0 No

Pop. variance [𝝈𝝈𝟐𝟐 ] 𝑠𝑠𝑋𝑋2 𝐸𝐸�𝑠𝑠2𝑋𝑋 � − 𝜎𝜎2𝑋𝑋 = 0 Yes, if X normal

2 2
Mean Squared Error 𝑀𝑀𝑀𝑀𝑀𝑀�𝜃𝜃�𝑋𝑋 � = 𝐸𝐸 ��𝜃𝜃�𝑋𝑋 − 𝜃𝜃𝑋𝑋 � � = 𝑉𝑉𝑉𝑉𝑉𝑉�𝜃𝜃�𝑋𝑋 � + �𝐵𝐵𝐵𝐵𝐵𝐵𝐵𝐵�𝜃𝜃�𝑋𝑋 ��

SAMPLE PROP. T-
DISTRIBUTION CHI-SQUARE FISHER’S F
(Bernoulli) STUDENT
𝑥𝑥
𝑝𝑝(1 − 𝑝𝑝) (𝑛𝑛 − 1)𝑆𝑆 2 ~ 𝑡𝑡𝑛𝑛 𝑋𝑋1 ⁄𝑛𝑛1 𝑠𝑠𝑠𝑠𝑋𝑋21 �𝜎𝜎𝑋𝑋21
ESTIMATOR 𝑝𝑝̂ ~ 𝑁𝑁 �𝑝𝑝, � � 2
~ 𝑋𝑋𝑛𝑛−1 ~ 𝐹𝐹𝑛𝑛1,𝑛𝑛2 ~ 𝐹𝐹𝑛𝑛1−1,𝑛𝑛2−1
𝑛𝑛 𝜎𝜎 2 �𝑌𝑌�𝑛𝑛 𝑋𝑋2 ⁄𝑛𝑛2 𝑠𝑠𝑠𝑠𝑋𝑋21 �𝜎𝜎𝑋𝑋22

PIVOTAL BOUNDS
PARAMETER ASSUMPTIONS CONFIDENCE INTERVAL
QUANTITY (Upper/Lower)
𝑥𝑥̅ − 𝜇𝜇0 𝜎𝜎 𝜇𝜇 ≤ 𝑥𝑥̅ + 𝑧𝑧𝛼𝛼 𝜎𝜎�
Normal data, ~ 𝑁𝑁(0,1) 𝜇𝜇 ∈ �𝑥𝑥̅ ∓ 𝑧𝑧𝛼𝛼�2 � √𝑛𝑛
𝜎𝜎
𝜎𝜎 known � 𝑛𝑛 √𝑛𝑛 𝜇𝜇 ≥ 𝑥𝑥̅ − 𝑧𝑧𝛼𝛼 𝜎𝜎�
√ √𝑛𝑛
𝑥𝑥̅ − 𝜇𝜇0 𝑠𝑠 𝜇𝜇 ≤ 𝑥𝑥̅ + 𝑧𝑧𝛼𝛼 𝑠𝑠�
Non-normal data, ~ 𝑁𝑁(0,1) 𝜇𝜇 ∈ �𝑥𝑥̅ ∓ 𝑧𝑧𝛼𝛼�2 � √𝑛𝑛
𝑠𝑠
𝑛𝑛 ≥ 30 � 𝑛𝑛 √𝑛𝑛 𝜇𝜇 ≥ 𝑥𝑥̅ − 𝑧𝑧𝛼𝛼 𝑠𝑠�
√ √𝑛𝑛
Population
𝑝𝑝̂ (1 − 𝑝𝑝̂ )
mean [𝝁𝝁] 𝑝𝑝̂ − 𝑝𝑝0 𝑝𝑝 ≤ 𝑝𝑝̂ + 𝑧𝑧𝛼𝛼 �
Bernoulli distrib., ~ 𝑁𝑁(0,1) 𝑝𝑝̂ (1 − 𝑝𝑝̂ ) 𝑛𝑛
𝑝𝑝 ∈ �𝑝𝑝̂ ∓ 𝑧𝑧𝛼𝛼�2 � �
𝑛𝑛 ≥ 30 �𝑝𝑝(1 − 𝑝𝑝) 𝑛𝑛 𝑝𝑝̂ (1 − 𝑝𝑝̂ )
𝑛𝑛 𝑝𝑝 ≥ 𝑝𝑝̂ − 𝑧𝑧𝛼𝛼 �
𝑛𝑛

𝑥𝑥̅ − 𝜇𝜇0 𝑠𝑠 𝜇𝜇 ≤ 𝑥𝑥̅ + 𝑡𝑡𝑛𝑛−1,𝛼𝛼 𝑠𝑠�


Normal data, ~ 𝑡𝑡𝑛𝑛−1 𝜇𝜇 ∈ �𝑥𝑥̅ ∓ 𝑡𝑡𝑛𝑛−1,𝛼𝛼�2 � √𝑛𝑛
𝑠𝑠 𝜇𝜇 ≥ 𝑥𝑥̅ − 𝑡𝑡𝑛𝑛−1,𝛼𝛼 𝑠𝑠�
𝜎𝜎 unknown � 𝑛𝑛 √𝑛𝑛
√ √𝑛𝑛
(𝑛𝑛 − 1)𝑠𝑠 2
2 2 𝜎𝜎 2 ≤
Variance (𝑛𝑛 − 1)𝑠𝑠 2 (𝑛𝑛 − 1)𝑠𝑠 (𝑛𝑛 − 1)𝑠𝑠 2
𝑋𝑋𝑛𝑛−1;1−𝛼𝛼
Normal data 2
~ 𝑋𝑋𝑛𝑛−1 𝜎𝜎 2 ∈ � , �
[𝝈𝝈𝟐𝟐 ] 𝜎𝜎 2
2
𝑋𝑋𝑛𝑛−1,𝛼𝛼�
2
𝑋𝑋𝑛𝑛−1,1− 𝛼𝛼� (𝑛𝑛 − 1)𝑠𝑠 2
2 2 𝜎𝜎 2 ≥ 2
𝑋𝑋𝑛𝑛−1;𝛼𝛼
(𝑛𝑛 − 1)𝑠𝑠 2
Standard 𝜎𝜎 ≤ � 2
(𝑛𝑛 − 1)𝑠𝑠 2 (𝑛𝑛 − 1)𝑠𝑠 2 (𝑛𝑛 − 1)𝑠𝑠 2 𝑋𝑋𝑛𝑛−1;1−𝛼𝛼
dev. Normal data 2
~ 𝑋𝑋𝑛𝑛−1 𝜎𝜎 ∈ �� 2 ,� 2 �
𝜎𝜎 2 𝑋𝑋𝑛𝑛−1,𝛼𝛼� 𝑋𝑋𝑛𝑛−1,1− 𝛼𝛼� (𝑛𝑛 − 1)𝑠𝑠 2
[𝒔𝒔. 𝒅𝒅. ] 2 2 𝜎𝜎 ≥ � 2
𝑋𝑋𝑛𝑛−1;𝛼𝛼
σ known z-based
(exact)
X ⁓ Normal
σ unknown
t-based
X ⁓ Distribution (exact)
µ, σ z-based
n large
(approx.)
X ≁ Normal
Methods
n small beyond Est II

ACTUAL SITUATION
DECISION
𝐻𝐻0 is TRUE 𝐻𝐻0 is FALSE
No error Type II error (𝛽𝛽)
ACCEPT 𝐻𝐻0
(1 − 𝛼𝛼) 𝑃𝑃(𝑓𝑓𝑓𝑓𝑓𝑓𝑓𝑓 𝑡𝑡𝑡𝑡 𝑟𝑟𝑟𝑟𝑟𝑟𝑟𝑟𝑟𝑟𝑟𝑟 𝐻𝐻0 |𝐻𝐻0 𝑖𝑖𝑖𝑖 𝑓𝑓𝑓𝑓𝑓𝑓𝑓𝑓𝑓𝑓)

REJECT 𝐻𝐻0 Type I error (𝛼𝛼) No error (1 − 𝛽𝛽 = 𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝)


𝑃𝑃(𝑟𝑟𝑟𝑟𝑟𝑟𝑟𝑟𝑟𝑟𝑟𝑟 𝐻𝐻0 |𝐻𝐻0 𝑖𝑖𝑖𝑖 𝑡𝑡𝑡𝑡𝑡𝑡𝑡𝑡) 𝑃𝑃(𝑟𝑟𝑟𝑟𝑟𝑟𝑟𝑟𝑟𝑟𝑟𝑟 𝐻𝐻0 |𝐻𝐻0 𝑖𝑖𝑖𝑖 𝑓𝑓𝑓𝑓𝑓𝑓𝑓𝑓𝑓𝑓)

p-value: The smallest value of 𝛼𝛼 for which 𝐻𝐻0 can be rejected.


 If p-value < 𝛼𝛼, reject 𝐻𝐻0 .
 If p-value ≥ 𝛼𝛼 , accept 𝐻𝐻0 .
 Small p-value → Evidence against 𝐻𝐻0 .
 Large p-value → Evidence in favour of 𝐻𝐻0 .

REJECTION REGION (RRα): �𝑡𝑡𝑡𝑡𝑡𝑡𝑡𝑡 > 𝑧𝑧𝛼𝛼�2 𝑜𝑜𝑜𝑜 𝑡𝑡𝑡𝑡𝑡𝑡𝑡𝑡 < −𝑧𝑧𝛼𝛼�2 � , �𝑡𝑡𝑡𝑡𝑡𝑡𝑡𝑡 ∉ [−𝑧𝑧𝛼𝛼�2 , 𝑧𝑧𝛼𝛼�2 ]�

PIVOTAL
PARAMETER ASSUMPTIONS REJECTION REGION IN TWO-TAIL TEST (RRα)
QUANTITY

Normal data, 𝑥𝑥̅ − 𝜇𝜇0 𝑥𝑥� − 𝜇𝜇 𝑥𝑥� − 𝜇𝜇


𝜎𝜎 ~ 𝑁𝑁(0,1) �𝑧𝑧: 𝜎𝜎 0 < 𝑧𝑧1−𝛼𝛼�2 𝑜𝑜𝑜𝑜 𝜎𝜎 0 > 𝑧𝑧𝛼𝛼�2 �
𝜎𝜎 known � 𝑛𝑛 � 𝑛𝑛 � 𝑛𝑛
√ √ √

Non-normal data, 𝑥𝑥̅ − 𝜇𝜇0 𝑥𝑥� − 𝜇𝜇0 𝑥𝑥� − 𝜇𝜇


𝑠𝑠 ~ 𝑁𝑁(0,1) �𝑧𝑧: 𝑠𝑠 < 𝑧𝑧1−𝛼𝛼�2 𝑜𝑜𝑜𝑜 𝑠𝑠 0 > 𝑧𝑧𝛼𝛼�2 �
𝑛𝑛 ≥ 30 � 𝑛𝑛 � 𝑛𝑛 � 𝑛𝑛
√ √ √
Population
mean [𝝁𝝁] 𝑝𝑝̂ − 𝑝𝑝0 ⎧ 𝑝𝑝
� − 𝑝𝑝0 𝑝𝑝
� − 𝑝𝑝0 ⎫
Bernoulli distrib., ~ 𝑁𝑁(0,1)
𝑧𝑧: < 𝑧𝑧1−𝛼𝛼�2 𝑜𝑜𝑜𝑜 > 𝑧𝑧𝛼𝛼�2
𝑛𝑛 ≥ 30 �𝑝𝑝(1 − 𝑝𝑝) ⎨
�𝑝𝑝(1 − 𝑝𝑝) �𝑝𝑝(1 − 𝑝𝑝)
⎬
𝑛𝑛 ⎩ 𝑛𝑛 𝑛𝑛 ⎭

Normal data, 𝑥𝑥̅ − 𝜇𝜇0 𝑥𝑥� − 𝜇𝜇0 𝑥𝑥� − 𝜇𝜇


𝑠𝑠 ~ 𝑡𝑡𝑛𝑛−1 �𝑡𝑡: 𝑠𝑠 < 𝑡𝑡𝑛𝑛−1; 1−𝛼𝛼�2 𝑜𝑜𝑜𝑜 𝑠𝑠 0 > 𝑡𝑡𝑛𝑛−1; 𝛼𝛼�2 �
𝜎𝜎 unknown � 𝑛𝑛 � 𝑛𝑛 � 𝑛𝑛
√ √ √

(𝑛𝑛 − 1)𝑠𝑠 2 (𝑛𝑛 − 1)𝑠𝑠2 (𝑛𝑛 − 1)𝑠𝑠2


Variance [𝝈𝝈𝟐𝟐 ] Normal data 2
~ 𝑋𝑋𝑛𝑛−1 �𝑋𝑋 2 : 2
< 𝑋𝑋𝑛𝑛−1; 1−𝛼𝛼� 𝑜𝑜𝑜𝑜
2
> 𝑋𝑋𝑛𝑛−1;𝛼𝛼� �
𝜎𝜎 2 𝜎𝜎2 2 𝜎𝜎2 2

Standard dev. (𝑛𝑛 − 1)𝑠𝑠 2 (𝑛𝑛 − 1)𝑠𝑠2 2


(𝑛𝑛 − 1)𝑠𝑠2 2
Normal data 2
~ 𝑋𝑋𝑛𝑛−1 �𝑋𝑋 2 : < 𝑋𝑋𝑛𝑛−1; 1−𝛼𝛼� 𝑜𝑜𝑜𝑜 > 𝑋𝑋𝑛𝑛−1;𝛼𝛼� �
[𝒔𝒔. 𝒅𝒅. ] 𝜎𝜎 2 𝜎𝜎2 2 𝜎𝜎2 2
PARAMETER ASSUMPTIONS PIVOTAL QUANTITY CONFIDENCE INTERVAL

� − 𝐷𝐷0
𝐷𝐷 𝑠𝑠𝐷𝐷
Normal differences ~ 𝑡𝑡𝑛𝑛−1 𝜇𝜇𝑋𝑋 − 𝜇𝜇𝑌𝑌 ∈ �𝑑𝑑̅ ∓ 𝑡𝑡𝑛𝑛−1,𝛼𝛼�2 �
𝑠𝑠𝐷𝐷
𝜎𝜎1 , 𝜎𝜎2 known � √𝑛𝑛
√𝑛𝑛
𝑋𝑋� − 𝑌𝑌� − 𝐷𝐷0
Normal pops. ~ 𝑡𝑡𝑛𝑛1 +𝑛𝑛2 −2 1 1
𝜇𝜇𝑋𝑋 − 𝜇𝜇𝑌𝑌 ∈ �𝑥𝑥̅ − 𝑦𝑦� ∓ 𝑡𝑡𝑛𝑛1 +𝑛𝑛2−2,𝛼𝛼�2 𝑆𝑆𝑝𝑝 � + �
𝜎𝜎12 = 𝜎𝜎22 𝑆𝑆𝑆𝑆�1�𝑛𝑛1 + 1�𝑛𝑛2 𝑛𝑛1 𝑛𝑛2

𝝁𝝁𝑿𝑿 − 𝝁𝝁𝒀𝒀 = 𝑫𝑫𝟎𝟎


𝑋𝑋� − 𝑌𝑌� − 𝐷𝐷0
Normal pops. ~ 𝑁𝑁(0,1) 𝜎𝜎2𝑋𝑋 𝜎𝜎2𝑌𝑌
2
𝜎𝜎𝑌𝑌2� 𝜇𝜇𝑋𝑋 − 𝜇𝜇𝑌𝑌 ∈ �𝑥𝑥̅ − 𝑦𝑦� ∓ 𝑧𝑧𝛼𝛼�2 � + �
𝜎𝜎1 , 𝜎𝜎2 known �𝜎𝜎𝑋𝑋� 𝑛𝑛1 + 𝑛𝑛2 𝑛𝑛1 𝑛𝑛2

Non-normal data 𝑋𝑋� − 𝑌𝑌� − 𝐷𝐷0


≈ 𝑁𝑁(0,1) 𝑠𝑠𝑋𝑋2 𝑠𝑠𝑌𝑌2
𝜎𝜎1 , 𝜎𝜎2 unknown 2
𝑠𝑠𝑌𝑌2� 𝜇𝜇𝑋𝑋 − 𝜇𝜇𝑌𝑌 ∈ �𝑥𝑥̅ − 𝑦𝑦� ∓ 𝑧𝑧𝛼𝛼�2 � + �
�𝑠𝑠𝑋𝑋� 𝑛𝑛1 + 𝑛𝑛2
𝑛𝑛1 𝑛𝑛2
Large samples

𝑝𝑝̂𝑋𝑋 − 𝑝𝑝̂ 𝑌𝑌
Bernoulli pops. ≈ 𝑁𝑁(0,1) 1 1
𝒑𝒑𝑿𝑿 − 𝒑𝒑𝒀𝒀 = 𝟎𝟎 1 1 𝑝𝑝𝑋𝑋 − 𝑝𝑝𝑌𝑌 ∈ �𝑝𝑝̂𝑋𝑋 − 𝑝𝑝̂𝑌𝑌 ∓ 𝑧𝑧𝛼𝛼�2 �𝑝𝑝̂ 0 (1 − 𝑝𝑝̂0 ) � + ��
Large samples �𝑝𝑝̂ 0 (1 − 𝑝𝑝̂ 0 ) �
𝑛𝑛1 + 𝑛𝑛2 �
𝑛𝑛1 𝑛𝑛2

𝝈𝝈𝟐𝟐𝑿𝑿 𝑠𝑠𝑋𝑋2 𝜎𝜎𝑋𝑋2 𝑠𝑠𝑋𝑋2 1 𝑠𝑠𝑋𝑋2 1


� 𝟐𝟐 = 𝟏𝟏 Normal pops. ~𝐹𝐹 ∈ � , �
𝝈𝝈𝒀𝒀 𝑠𝑠𝑌𝑌2 𝑛𝑛1−1,𝑛𝑛2−1 𝜎𝜎𝑌𝑌2 𝑠𝑠𝑌𝑌2 𝐹𝐹𝑛𝑛1−1,𝑛𝑛2−1,𝛼𝛼 𝑠𝑠𝑌𝑌2 𝐹𝐹𝑛𝑛1−1,𝑛𝑛2−1,1−𝛼𝛼

(𝑛𝑛1 − 1)𝑠𝑠𝑋𝑋2 + (𝑛𝑛2 − 1)𝑠𝑠𝑌𝑌2 𝑛𝑛1 𝑝𝑝̂𝑋𝑋 + 𝑛𝑛2 𝑝𝑝̂ 𝑌𝑌 1


𝑆𝑆𝑝𝑝2 = 𝑝𝑝̂ 0 = 𝐹𝐹𝑛𝑛.𝑚𝑚,𝛼𝛼 =
𝑛𝑛1 + 𝑛𝑛2 − 2 𝑛𝑛1 + 𝑛𝑛2 𝐹𝐹𝑚𝑚,𝑛𝑛,1−𝛼𝛼

𝐼𝐼𝐼𝐼 𝐷𝐷0 ∈ 𝐶𝐶𝐼𝐼1−𝛼𝛼 ⟹ 𝐹𝐹𝐹𝐹𝐹𝐹𝐹𝐹 𝑡𝑡𝑡𝑡 𝑟𝑟𝑟𝑟𝑟𝑟𝑟𝑟𝑟𝑟𝑟𝑟 𝐻𝐻0 𝐼𝐼𝐼𝐼 𝐷𝐷0 ∉ 𝐶𝐶𝐼𝐼1−𝛼𝛼 ⟹ 𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑅 𝐻𝐻0

PARAMETER CONFIDENCE INTERVAL � 𝟎𝟎 + 𝜷𝜷


� = 𝜷𝜷
𝒚𝒚 � 𝟏𝟏 𝒙𝒙

� 1 − 𝛽𝛽1
𝛽𝛽
𝑐𝑐𝑐𝑐𝑐𝑐(𝑥𝑥, 𝑦𝑦) 𝑆𝑆𝑅𝑅2 ~ 𝑡𝑡𝑛𝑛−2
𝛽𝛽̂1 = 𝛽𝛽1 ∈ �𝛽𝛽̂1 ∓ 𝑡𝑡𝑛𝑛−2,𝛼𝛼�2 � � 𝑠𝑠𝑅𝑅2
𝑠𝑠𝑋𝑋2 (𝑛𝑛 − 1)𝑠𝑠𝑋𝑋2 �
(𝑛𝑛 − 1)𝑠𝑠𝑋𝑋2

� 0 − 𝛽𝛽0
𝛽𝛽
1 𝑥𝑥̅ 2 ~ 𝑡𝑡𝑛𝑛−2
𝛽𝛽̂0 = 𝑦𝑦� − 𝛽𝛽̂1 𝑥𝑥̅ 𝛽𝛽0 ∈ �𝛽𝛽̂0 ∓ 𝑡𝑡𝑛𝑛−2,𝛼𝛼�2 �𝑆𝑆𝑅𝑅2 � + �� 1 �2
𝑥𝑥
𝑛𝑛 (𝑛𝑛 − 1)𝑠𝑠𝑋𝑋2 �𝑆𝑆2𝑅𝑅 � + �
𝑛𝑛 (𝑛𝑛 − 1)𝑠𝑠2𝑋𝑋

∑ 𝑒𝑒𝑖𝑖2
𝑆𝑆𝑅𝑅2 = (𝑛𝑛 − 2)𝑆𝑆𝑅𝑅2 (𝑛𝑛 − 2)𝑆𝑆𝑅𝑅2 𝑠𝑠𝑅𝑅2
𝑛𝑛 − 2 𝜎𝜎 2 ∈ � 2 , 2 � 𝑠𝑠�𝛽𝛽� 𝑖𝑖 � = �
𝜒𝜒𝑛𝑛−2, 𝛼𝛼� 𝜒𝜒𝑛𝑛−2,1−𝛼𝛼� (𝑛𝑛 − 1)𝑠𝑠𝑋𝑋2
Simple linear regression 2 2

1 (𝑥𝑥̅ − 𝑥𝑥)2
𝐸𝐸[𝑌𝑌0 ] ∈ �𝑌𝑌�0 ∓ 𝑡𝑡𝑛𝑛−2,𝛼𝛼�2 �𝑆𝑆𝑅𝑅2 � + ��
𝑛𝑛 (𝑛𝑛 − 1)𝑠𝑠𝑋𝑋2
� 𝑖𝑖 − 𝛽𝛽𝑖𝑖
𝛽𝛽
𝑦𝑦� = 𝑦𝑦� − 𝛽𝛽̂1 (𝑥𝑥̅ − 𝑥𝑥) ~ 𝑡𝑡𝑛𝑛−𝑘𝑘−1
𝑠𝑠�𝛽𝛽� 𝑖𝑖 �
1 (𝑥𝑥̅ − 𝑥𝑥)2
𝑌𝑌0 ∈ �𝑌𝑌�0 ∓ 𝑡𝑡𝑛𝑛−2,𝛼𝛼�2 �𝑆𝑆𝑅𝑅2 �1 + + ��
𝑛𝑛 (𝑛𝑛 − 1)𝑠𝑠𝑋𝑋2

∑ 𝑒𝑒𝑖𝑖2
𝑆𝑆𝑅𝑅2 =
𝑛𝑛 − 𝑘𝑘 − 1 𝛽𝛽𝑖𝑖 ∈ �𝛽𝛽̂𝑖𝑖 ± 𝑡𝑡𝑛𝑛−𝑘𝑘−1,𝛼𝛼�2 𝑠𝑠�𝛽𝛽̂𝑖𝑖 �� 𝛽𝛽̂ = (𝑋𝑋 𝑇𝑇 𝑋𝑋)−1 𝑋𝑋 𝑇𝑇 𝑌𝑌 𝑦𝑦� = 𝑋𝑋𝛽𝛽̂
Multiple linear regression
∑𝑥𝑥𝑖𝑖2 − 𝑛𝑛(𝑥𝑥̅ )2 Covariance ∑𝑥𝑥𝑖𝑖 𝑦𝑦𝑖𝑖 − 𝑛𝑛(𝑥𝑥̅ )(𝑦𝑦�)
Variance [𝒔𝒔𝟐𝟐𝒙𝒙𝒊𝒊 ]
𝑛𝑛 − 1 [𝒄𝒄𝒄𝒄𝒄𝒄(𝒙𝒙, 𝒚𝒚)] 𝑛𝑛 − 1

Correlation 𝑐𝑐𝑐𝑐𝑐𝑐(𝑥𝑥, 𝑦𝑦)


[𝒄𝒄𝒄𝒄𝒄𝒄𝒄𝒄(𝒙𝒙, 𝒚𝒚)] 𝑠𝑠𝑋𝑋 𝑠𝑠𝑌𝑌

SSM = Squared Sum Modal �(𝑦𝑦�𝑖𝑖 − 𝑦𝑦�)2

SSR = Squared Sum Residual �(𝑦𝑦�𝑖𝑖 − 𝑦𝑦𝑖𝑖 )2

SST = Squared Total Sum �(𝑦𝑦𝑖𝑖 − 𝑦𝑦�)2

𝑆𝑆𝑆𝑆𝑆𝑆 = 𝑆𝑆𝑆𝑆𝑆𝑆 + 𝑆𝑆𝑆𝑆𝑆𝑆


𝑆𝑆𝑆𝑆𝑆𝑆 𝑆𝑆𝑆𝑆𝑆𝑆
R2=r2=Coefficient of determination 𝑅𝑅2 = =1−
𝑆𝑆𝑆𝑆𝑆𝑆 𝑆𝑆𝑆𝑆𝑆𝑆

ANOVA table
Source of variability SS DF Mean F ratio
Model SSM 𝑘𝑘 𝑆𝑆𝑆𝑆𝑆𝑆/𝑘𝑘 (𝑆𝑆𝑆𝑆𝑆𝑆/𝑘𝑘)/𝑠𝑠𝑅𝑅2
Residuals/errors SSR 𝑛𝑛 − 𝑘𝑘 − 1 𝑆𝑆𝑆𝑆𝑆𝑆/(𝑛𝑛 − 𝑘𝑘 − 1)
Total SST 𝑛𝑛 − 1 𝑆𝑆𝑆𝑆𝑆𝑆/𝑘𝑘
𝐹𝐹 =
𝑠𝑠𝑅𝑅2

𝐻𝐻0 : 𝑝𝑝𝑖𝑖 = 𝑝𝑝𝑖𝑖0 ∀𝑖𝑖 ∈ {1, … , 𝑘𝑘} 𝐻𝐻1 : 𝑝𝑝𝑖𝑖 ≠ 𝑝𝑝𝑖𝑖0 ∀𝑖𝑖 ∈ {1, … , 𝑘𝑘}
2
�𝑁𝑁𝑖𝑖 − 𝑛𝑛𝑝𝑝𝑖𝑖0 � 2
𝑄𝑄 = 𝑄𝑄~ 𝜒𝜒𝑘𝑘−1
𝑛𝑛𝑝𝑝𝑖𝑖0
2
𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑅 𝐻𝐻0 𝑖𝑖𝑖𝑖 𝑄𝑄 > 𝜒𝜒𝑘𝑘−1,𝛼𝛼

You might also like