Scribd Logo
Upload

Welcome to Scribd!

  • 3 views

    Bio-L8- Correlation and Regression Analysis

    Uploaded by

    Saad

    Copyright:

    © All Rights Reserved
    Download as PDF, TXT or read online from Scribd

    Bio-L8- Correlation and Regression Analysis

    Uploaded by

    Saad
    3 views15 pages

    Copyright

    © © All Rights Reserved

    Available Formats

    PDF, TXT or read online from Scribd

    Did you find this document useful?

    Is this content inappropriate?

    Copyright:

    © All Rights Reserved
    Download as PDF, TXT or read online from Scribd
    Download as pdf or txt
    3 views15 pages

    Bio-L8- Correlation and Regression Analysis

    Uploaded by

    Saad

    Copyright:

    © All Rights Reserved
    Download as PDF, TXT or read online from Scribd
    Download as pdf or txt
    of 15

    Linear Regression and Correlation

    Correlation and Regression:


    It frequently happens that statistic want to discribe with a single number 𝑎
    relationship between two sets of scores. [ A number which measures a
    relationship between two sets of scores is called a correlation coefficient ].

    Scatter Gram :
    Consider a list of pairs of numerical values respresenting variables 𝑥 and 𝑦 .
    The scatter gram of the data is simply a picture of the pairs of values as a
    point in a coordinate plan 𝑅3 . The picture some times indicates a
    relationship between the point as illustrated in the following examples :
    Correlation Coefficient :
    Pearson defined 𝑟 so that the formula for 𝑟 has a minimum possible value of −1 and
    a maximum possible value of +1 , when the sample points lie exatly in a line sloping
    down to the right we say there is perfect negative correlation : 𝑟 = −1 , when the
    sample points lie exactly in a line sloping up to the right , we say there is perfect
    positive correlation : 𝑟 = +1 , when there is no tending of the points to the lie in a
    straight line we say there is no correlation 𝑟 = 0 .
    If 𝑟 is near to +1 or −1 we say we have high correlation. If 𝑟 is near zero, we say
    that we have lowe correlation .
    𝑛 𝑛
    𝑖=1(𝑥𝑖 −𝑋) 𝑖=1(𝑦𝑖 −𝑌)
    𝑟=
    𝑛 (𝑥 −𝑋)2 𝑛 (𝑦 −𝑌)2
    𝑖=1 𝑖 𝑖=1 𝑖

    or
    𝑛 𝑛 𝑛
    𝑖=1 𝑥𝑖 𝑦𝑖 − ( 𝑖=1 𝑥𝑖 𝑖=1 𝑦𝑖 )/𝑛
    𝑟=
    𝑛 2 𝑛 2
    𝑛 2
    ( 𝑖=1 𝑥𝑖 ) 𝑛 2
    ( 𝑖=1 𝑦𝑖 )
    𝑖=1 𝑥𝑖 − 𝑛 𝑖=1 𝑦𝑖 − 𝑛
    64 81 81
    Regression Line :
    A regression line is a straight line that describes how a response variable
    y changes as an explanatory variable x changes. We often use a
    regression line to predict the value of y for a given value of x.

    𝑦 = 𝑎 + 𝑏𝑥

    𝑛
    𝑖=1 𝑥𝑖 𝑦𝑖 −𝑛𝑋𝑌
    𝑏= 𝑛 𝑥 2 −𝑛𝑋 2
    𝑖=1 𝑖

    𝑎 = 𝑌 − 𝑏𝑋
    𝑛 𝑛 𝑛
    𝑖=1 𝑥𝑖 𝑦𝑖 = 711 , 𝑖=1 𝑥𝑖 = 47 , 𝑋 = 47/6 , 𝑖=1 𝑦𝑖 = 79 , 𝑌 = 79/6

    𝑛 2
    𝑖=1 𝑥𝑖 = 423 ,

    𝑏 = 1.68 , 𝑎 = 𝑌 − 𝑏𝑋 = 0.005 , 𝑦 = 0.005 + 1.68𝑥

    b.) when 𝑥 = 4 , 𝑦 = 0.005 + 1.68 ∗ (4) = 6.725

    when 𝑥 = 1 , 𝑦 = 0.005 + 1.68 ∗ (1) = 1.685

    when 𝑥 = 15 , 𝑦 = 0.005 + 1.68 ∗ 15 = 25.205

    Example : let we have the following data :


    𝑥 6 6 7 8 9 9
    𝑦 5 6 6 7 7 8

    then , 1) construct a scatter gram and draw a calculated regression line .

    2) find the correlation coefficient and explain it .


    Simple and Multiple Linear Regression Models
    Definition: A multiple linear regression model
    relating a random response 𝑌 to a set of predictor
    variables 𝑥1 , . . . , 𝑥𝑛 is an equation of the form
    𝑌 = 𝛽0 + 𝛽1 𝑥1 + 𝛽2 𝑥2 +· · ·+𝛽𝑘 𝑥𝑘 + ε
    Where 𝛽0 , . . . , 𝛽𝑘 are unknown parameters,
    𝑥1 , . . . , 𝑥𝑘 are the independent non-random
    variables, and 𝜀 is a random variable
    representing an error term. We assume that
    𝐸(𝜀) = 0, or equivalently
    , 𝐸(𝑌) = 𝛽0 + 𝛽1 𝑥1 + 𝛽2 𝑥2 +· · ·+𝛽𝑘 𝑥𝑘 .

    Definition: If 𝑌 = 𝛽0 + 𝛽1 𝑥+𝜀 , this


    is called a simple
    linear regression model. Here, 𝛽0 is
    the y-intercept of the line and 𝛽1 is
    the slope of the line. The term ε is
    the error component.
    11
    The Method of Least Squares
    Let (𝑥1 , 𝑦1 ), ( 𝑥2 , 𝑦2 ), . . . ( 𝑥𝑛 , 𝑦𝑛 ), are the 𝑛 observed data points, with
    corresponding errors 𝜀𝑖 , 𝑖 = 1, . . . , 𝑛. That is, 𝑌𝑖 = 𝛽0 + 𝛽1 𝑥𝑖 +𝜀𝑖 , 𝑖 = 1, 2, . . . , 𝑛.
    We assume that the errors 𝜀𝑖 , 𝑖 = 1, 2, . . . , 𝑛 are independent and identically
    distributed with 𝐸(𝜀𝑖 ) = 0 , 𝑖 = 1, 2, . . . , 𝑛 and 𝑉𝑎𝑟(𝜀𝑖 ) = 𝜎 2 , 𝑖 = 1, 2, . . . , 𝑛 .
    One of the ways to decide on how well a straight line fits the set of data is to
    determine the extent to which the data points deviate from the line. The
    straight line model for the response 𝑌 for a given 𝑥 is 𝑌 = 𝛽0 + 𝛽1 𝑥+𝜀 .
    Because we assumed that 𝐸(𝜀) = 0, the expected value of 𝑌 is given by
    𝐸(𝑌) = 𝛽0 + 𝛽1 𝑥.
    The estimator of the 𝐸(𝑌), denoted by 𝑌, can be obtained by using the
    estimators 𝛽0 and 𝛽1 of the parameters 𝛽0 and 𝛽1 , respectively. Then, the fitted
    regression line we are looking for is given by 𝑌 = 𝛽0 + 𝛽1 𝑥.
    For observed values (𝑥𝑖 , 𝑦𝑖 ),we obtain the estimated value of 𝑦𝑖 as
    𝑦𝑖 = 𝛽0 + 𝛽1 𝑥𝑖 .
    12
    The deviation of observed 𝑦𝑖 from its predicted value 𝑦𝑖 , called the 𝑖𝑡ℎ residual, is
    defined by 𝑒𝑖 = 𝑦𝑖 − 𝑦𝑖 = [𝑦𝑖 − (𝛽0 + 𝛽1 𝑥𝑖 )].
    The residuals, or errors 𝑒𝑖 , are the vertical distances between observed and predicted
    values of 𝑦𝑖 ’s.
    Definition: The sum of squares for errors (SSE) or sum of squares of the residuals for
    all of the 𝑛 data points is 𝑆𝑆𝐸 = 𝑛𝑖=1 𝑒𝑖2 = 𝑛𝑖=1[𝑦𝑖 −(𝛽0 + 𝛽1 𝑥𝑖 )]2 .
    The least-squares approach to estimation is to find 𝛽0 and 𝛽1 that minimize the sum
    of squared residuals, SSE.
    Derivation of 𝜷𝟎 and 𝜷𝟏 : To simplify the formula for 𝛽1 , set
    𝑛 2 ( 𝑛
    𝑖=1 𝑥𝑖 )
    2
    𝑛
    𝑛
    𝑖=1 𝑥𝑖
    𝑛
    𝑖=1 𝑦𝑖
    𝑆𝑥𝑥 = 𝑖=1 𝑥𝑖 − , 𝑆𝑥𝑦 = 𝑖=1 𝑥𝑖 𝑦𝑖 −
    𝑛 𝑛
    𝑆𝑥𝑦
    𝛽1 =  𝛽0 = 𝑦 − 𝛽1 𝑥  𝑌 = 𝛽0 + 𝛽1 𝑥.
    𝑆𝑥𝑥

    13
    Example 1:Use the method of least squares to fit a straight line to the accompanying data points.
    Give the estimates of 𝛽0 and 𝛽1 . Plot the points and sketch the fitted least-squares line. The observed
    data values are given in the following table. 𝑥 −1 0 2 −2 5 6 8 11 12 −3
    𝑦 −5 −4 2 −7 6 9 13 21 20 −9

    Solution: Form a table to compute various terms

    14
    𝑛 2 ( 𝑛
    𝑖=1 𝑥𝑖 )
    2 (38)2
    𝑆𝑥𝑥 = 𝑖=1 𝑥𝑖 − 𝑛
    = 408 − 10
    = 263.6
    𝑛 𝑛
    𝑛 𝑖=1 𝑥𝑖 𝑖=1 𝑦𝑖 38 46
    𝑆𝑥𝑦 = 𝑖=1 𝑥𝑖 𝑦𝑖 − 𝑛
    = 709 − 10
    = 534.2

    𝑥 = 3.8, 𝑦 = 4.6
    𝑆𝑥𝑦 534.2
    Therefore, 𝛽1 = 𝑆𝑥𝑥 = 263.6 = 2.0266
    and 𝛽0 = 𝑦 − 𝛽1 𝑥 = 4.6 − (2.0266)(3.8) = −3.1011
    Hence, the least-squares line for these data is
    𝑦𝑖 = 𝛽0 + 𝛽1 𝑥𝑖 = −3.1011 + 2.0266 x

    EXAMPLE : Fit a least square line for the following data. Also find the trend values
    (𝑦 ) and show that (𝑦 − 𝑦)=0

    𝑥 1 2 3 4 5
    𝑦 2 5 3 8 7
    H.W

    15

    You might also like