Scribd Logo
Upload

Welcome to Scribd!

  • 375 views

    Path Analysis For Linear Multiple Regression

    Uploaded by

    Abhijit Kar Gupta
    A quick introduction to Path Analysis with an example (for linear multiple regression). This may be useful for the students of Geography and other social sciences.

    Copyright:

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

    Path Analysis For Linear Multiple Regression

    Uploaded by

    Abhijit Kar Gupta
    375 views3 pages
    A quick introduction to Path Analysis with an example (for linear multiple regression). This may be useful for the students of Geography and other social sciences.

    Original Title

    Path Analysis for Linear Multiple Regression

    Copyright

    © © All Rights Reserved

    Available Formats

    PDF, TXT or read online from Scribd

    Did you find this document useful?

    Is this content inappropriate?

    A quick introduction to Path Analysis with an example (for linear multiple regression). This may be useful for the students of Geography and other social sciences.

    Copyright:

    © All Rights Reserved
    Download as PDF, TXT or read online from Scribd
    Download as pdf or txt
    375 views3 pages

    Path Analysis For Linear Multiple Regression

    Uploaded by

    Abhijit Kar Gupta
    A quick introduction to Path Analysis with an example (for linear multiple regression). This may be useful for the students of Geography and other social sciences.

    Copyright:

    © All Rights Reserved
    Download as PDF, TXT or read online from Scribd
    Download as pdf or txt
    of 3
    At a glance
    Powered by AI
    Path analysis is a technique that uses path diagrams and regression analysis to study direct and indirect relationships between multiple variables. It allows us to decompose correlation coefficients to separate direct and indirect effects.

    Path analysis is an extension of multiple linear regression that is used to model causal relationships between variables. It uses path diagrams with single-headed arrows to represent assumed causal relationships, and allows decomposition of total correlations between variables into direct and indirect effects.

    Path coefficients are calculated by setting up equations using correlation coefficients between variables. The correlation coefficient between two directly linked variables gives the path coefficient. The equations are then solved to determine all path coefficients.

    1

    PATH ANALYSIS
    From Multiple Regressions

    Path analysis is an extension of the linear multiple regression analysis. The output (Y) is dependent on
    two or more input variables (X1, X2). The method of Path analysis is basically a hypothesis based
    modeling to identify the cause and effect and to know how the variables are related directly and
    indirectly.
    For pictorial display, the arrows show the assumed causal relations between two quantities. A single
    headed arrow (
    points from cause to effect. A double headed arrow (
    or a double headed curved
    arrow indicates that the variables are merely correlated but we are not sure about their causal relation.
    The independent variables (X1, X2) are called exogenous variables. The dependent (Y) variables are
    called endogenous variables.
    A path from 1 to 2 is designated by the path coefficient,
    effect (2) is listed first and then the cause (1).].

    [Path to 2 from 1. Note: In the suffix, the

    Let us suppose, we have the variable, Y dependent on X1 and X2. If we know that X1 and X2 are also
    related to each other, we can think of a model. To illustrate, we designate X1, X2 and Y as 1, 2, 3
    respectively. In our model, we can think that X2 is dependent on X1 and Y is dependent on X1 and X2.
    So we draw the path analysis diagram:
    The path coefficients between 1
    and 3 is , the path coefficient
    between 1 and 2 is and
    between 2 and 3 is

    1
    3
    2

    Now we see there is no path towards 1, one path from 1 to 2 and two paths towards 3 (from 1 and 2).
    Now if we know the correlation coefficients between all of them ( , ,
    that we prepared while doing multiple regression analysis, we can write,

    ) from the correlation table

    (1)
    (2)
    (3)
    Equation (1) tells us that there is a direct and only one path from 1 to 2 and so the path coefficient is the
    correlation coefficient between 1 & 2.

    Application of Statistics for Geography and other social sciences


    by Dr. Abhijit Kar Gupta. [email protected]

    2
    Equation (2), there are two contributions:
    is the direct path from 1 to 3 plus there is one path from 2
    to 3 ( ) but as 2 is correlated to , we multiply that by .
    Similarly, in equation (3), the correlation between 2 & 3 has two parts: direct path (
    coming from 1 and now we consider 1 is correlated to 2.

    ) plus the path

    Now consider the following multiple regression problem:


    20

    15

    20

    13

    18

    14

    28

    16

    24

    20

    23

    25

    12

    15

    12

    260

    240

    245

    210

    230

    242

    295

    235

    287

    252

    270

    275

    We calculated,

    . Here,

    To find correlation coefficients between each pair of variables, we need to first calculate the variances
    and covariances.

    )(

    )(

    )(

    NOTE: To remember the covariance formula, we can think of variance,

    )(

    ), as if it is a covariance between and . Thus

    following this we can write, covariance between and ,

    Correlation coefficients:
    Application of Statistics for Geography and other social sciences
    by Dr. Abhijit Kar Gupta. [email protected]

    )(

    ).

    So we can prepare a correlation Matrix:

    -0.81

    0.87

    -0.93
    1

    Now if we call
    and

    ,
    .

    and

    as 1, 2 and 3 for our purpose, we now have

    So from equations, (1), (2) and (3), we can write,


    ,

    Note that the path analysis depends on the


    choice of the model. We here analysed for a
    particular model. You can have various other
    models, for example, X1 can be dependent on
    X2 (rather than the opposite) or there may be
    no dependence at all. Calculations will be
    different.

    (4)
    (

    (5), and
    (

    (6)

    Solving the last two equations, (5) & (6) by multiplying (5) by 0.81 and adding with (6) etc., we get
    and
    So, now we have determined the path coefficients and now we can analyze the direct and indirect
    effects of one variable on the other. The path diagram with coefficients:

    X1

    So we see, there is the direct effect of X1 on Y = 0.34

    0.34

    -0.81

    X2

    The indirect effect of X1 on Y through X2 = (Effect of X1 on


    (
    X2)x(Effect of X2 on Y) = (
    So we conclude that the indirect effect is much more than
    the direct effect!

    -0.66

    Application of Statistics for Geography and other social sciences


    by Dr. Abhijit Kar Gupta. [email protected]

    You might also like