Mediational analysis

• Most research focuses on relations between two variables, X and Y, and much has been
written about two-variable relations, including conditions under which X can be considered
a possible cause of Y.
• Mediation in its simplest form represents the addition of a third variable to this X → Y
relation, whereby X causes the mediator, M, and M causes Y, so X → M → Y. Mediation is
only one of several relations that may be present when a third variable, Z (using Z to
represent the third variable), is included in the analysis of a two-variable system. This type
of relationship is called mediation, and the Z variable is called a mediator or mediating
variable (M).
It has two commonly used approaches:
Partial mediation: Partial mediation occurs when the mediating variable accounts for some,
but not all, of the relationship between the IV and the DV. In this case, the direct effect of the
IV on the DV remains significant even after accounting for the mediating variable. It suggests
that the mediating variable partially explains the relationship between the IV and the DV. This
implies that there may be other factors or pathways at play in influencing the DV.
Complete Mediation: Complete mediation occurs when the mediating variable fully explains
the relationship between the IV and the DV. In this scenario, the direct effect of the IV on the
DV becomes non-significant or reduced to zero when the mediating variable is included in
the analysis. The mediating variable serves as the sole explanatory mechanism through
which the IV affects the DV.

REGRESSION ANALYSIS
• Regression analysis is a statistical method used to examine the relationship between a
dependent variable and one or more independent variables.
• Purpose of regression analysis to understand how changes in Independent variable
affect the Dependent variable.
The type of regression analysis we are focusing on is simple linear regression.
• It consist of a regression line, also known as line of best fit, it is a straight line that
represents the relationship between two variable in regression analysis.
• The line of best fit represents the average relationship between two variables
 For eg.
Y-Values
Hours studies (X) Exam scores (Y) 90 85
80
2 65 80
70
75
70 65
3 70
60
4 75 50
40
5 80 30
6 85 20
10
0
Here is small sample of the data. 1.5 2 2.5 3 3.5 4 4.5 5 5.5 6 6.5

Hours studies (X)

• Lets consider this example to understand how line of best fit shows the relationship
between variables. To find the regression line, we perform linear regression analysis on this
data.
• The equation depicting this relationship would be the regression line equation i.e.
Y= a + bX
where, Y represents the dependent variable (exam scores) and X is the independent variable
(hour studies).
 The regression analysis estimates the values of a (intercept) and b (slope).
Assume the values of a and b, 55 and 5 respectively.
• The estimated regression equation would be :
Y= 55 + 5X
• This represents that when a student doesn’t study (X=0), the predicted exam score (Y)
would be 55.
• And for each additional hour of study, the predicted exam score is expected to increase by 5
points.
 That was about regression analysis that plays a significant role in understanding the
working of mediational analysis.
SINGLE MEDIATOR MODEL: A single mediator model is a statistical framework used to
analyze the relationship between a single independent variable (IV) and a single dependent
variable (DV) while considering the mediating effect of a single mediator variable. It aims to
understand how the IV influences the DV through the mediator, providing insights into the
underlying mechanisms and processes of the relationship.
 e3

Mediator (M)

a
b
e1 e2

Independent Dependent Independent Dependent

variable (X) variable (Y) variable (X) variable (Y)
C e1 c

Path diagram and equation for regression model. Path diagram and equation for mediational model

Y= i1+CX+e1 Y= i2+cX+bM+e2

M= i3+aX+e3
To conduct a single mediational analysis, we would typically use a regression-based
approach. Three regression equation has been used to investigate mediation. It basically
involves three steps suggested by the Baron and Kenny (1986).

First equation,
Y= i1+CX+e1
This equation is representing the direct relationship between independent variable (X) and
dependent variable(Y) where, C is the parameter relating the X and Y, whereas i1 and e1
represents the intercept and error variability respectively.

Second equation,
Y= i2+cX+bM+e2
This equation represents mediating effect of M on the relationship between X and Y. Where c is the
parameter relating X to Y adjusting for the effect of the mediator and b is the parameter relating
the mediator to Y adjusting for the effects of Y.

Third equation,
M= i3+aX+e3
This equation represents the relationship of mediator(M) to the independent variable(X).
Direct and indirect effect
• Direct effect refers to the relationship between X and Y that operates
independently of the mediating variable M. That could be seen in equation 1.
• Indirect effect refers to the relationship between X and Y that operates through
the mediating variable M. The mediated effect in the single-mediator model may
be calculated in two ways, as either ab or ĉ − ĉ′.

M Di
e ct re
ct
t eff eff
c ec
ire t
D

X Y
Direct effect

Indirect effect (a*b)

• The product of coefficients method, involves computing the product
of a and b, ab, to form the mediated or indirect effect.
• After computing the indirect effect. Asses the significance of the indirect effect
that can be estimated by computing the product is then divided by the standard
error of the product and the ratio is compared to a standard normal distribution.
• Significance can also be assess by two more approaches i.e. Sobel test and
bootstrapping.
• If the indirect effect is statistically significant, it suggests that M partially mediates
the relationship between X and Y.

ASSUMPTIONS OF SINGLE MEDIATIONAL MODEL

• Temporal Order: The assumption of temporal order suggests that the independent
variable (X) precedes the mediating variable (M), and M precedes the dependent
variable (Y) in time. This assumption is crucial in establishing a causal relationship
between X, M, and Y. Without a proper temporal sequence, it becomes challenging
to establish a mediating effect.
• No Measurement Error: The assumption of no measurement error implies that the
variables are measured without any systematic error or bias. Accurate and reliable
measurement of X, M, and Y is crucial to obtain valid results in mediational
analysis. Measurement error can introduce noise and attenuate the estimated
effects.
• Absence of Omitted Variables: The assumption of absence of omitted variables
implies that there are no unaccounted variables that directly influence both the
mediating variable (M) and the dependent variable (Y) without being included in
the analysis. Omitted variables could lead to biased estimates of the mediating
effect.