Path Analysis

3.11.2 Path Analysis

  Path analysis is used to describe and test the relationship between variables in the form of causation, thus in the relationship model among variables, there are independent and dependent variables (Sugiyono,2012). Path analysis also a model of a Path analysis is used to describe and test the relationship between variables in the form of causation, thus in the relationship model among variables, there are independent and dependent variables (Sugiyono,2012). Path analysis also a model of a

  X and Y variables (Bungin, 2009). Path analysis model is useful to know the causal relationship; furthermore this path aims to find out the direct or indirect relationship of cause of some variables (exogenous) to the variable result (endogenous). Path analysis provides a framework for the researcher to think more carefully about how the X variable is related to the Y as well as the X variable is related to each other. Moreover, to know the relationship between X variable to Y variable with or without Z variable which a mediator.

  Stages in performing analysis using path analysis are (Solimun in Maharani, 2013):

  1. Designing models based on concepts and theories on path diagrams using two

  kinds of relationships that are direct and indirect:

  a. Direct states the influence of free variables (Store Online Atmosphere) to the dependent variable (Purchase Intention).

  b. Indirect expresses the indirect influence between the free variable (Store Online Atmosphere) on the bound variable (Purchase Intention) through intervening variable (Attitude Toward Website).

  Figure 3. 1 The Model of Line Diagram Hypothesis

  ε 1

  Attitude toward website (Z)

  Alpha(α) Beta (β)

  Online store

  Purchase intention

  Source: Researcher,2018

  2. Look for the influence directly and indirectly To seek influence directly between endogenous and exogenous variables, then it must be made in advance with appropriate structure equations of the flow chart contained in the line.

  a. Z = PX+ ε 1 (Substructure 1)

  b. Y = PX+ PZ + ε 1 (Substructure 2)

  Description:

  X = Store Online Atmosphere

  Y = Purchase Intention

  Z = Attitude Toward Website

  ε 1 = Residue Variable or variables that affect Y1, but not discussed in this study ε 2 = Residue Variable or variables that affect Y2, but not discussed in this study Logical steps to testing for a mediator:

  a. Test if there is significant relationship between X and Z (total causal effect of X on Z), record this effect

  b. Test if there is significant relationship between X and Y (Total causal effect of X on Y), record this effect

  c. Estimate the effect of X and Y have simultaneously on Z, record the Y and Z effect as β

  If is not significantly different than zero, then the causal effect of X on Z is fully mediated by Y. if < but is still statistically significant, then it could be say the

  relationship between X and Z is partially mediated by Y. Sometimes see

  presented as the percent of the total causal effect of X on Z explained by the mediator Y.

  4 Calculates the individual path coefficients

  Furthermore, to determine the significance of the path analysis, then the probability value of 0.05 is smaller or equal to the probability value of Sig or (0.05 Sig), then Ho is rejected and Ha is accepted, meaning significant.

  5 Summarize

  Summarize the results of research and compare with the results of previous research. Finally, it is continued by summarizing the results of the study as a whole and provide good advice for further researchers and research objects.

  3.12 Hypothesis Testing

  According Sugiyono (2014) hypothesis is defined as a temporary answer to the formulation of research problems. The truth of the hypothesis must be proven through the collected data. Hypothesis testing is a procedure used to test the truth of a statement statistically so that it can be deduced to decide whether the statement is rejected or accepted. This test is done by comparing t count and t table with significance level t<0,05 (5) and at free degree degree df = n-k-l. If t count >t tabel then the independent variable has no effect on the dependent variable. Similarly if the significance of F under 0.05 then simultaneously independent variables affect the dependent variable, and vice versa if the significance is above 0.05 then the independent variable has no effect on the dependent variable.

  3.13 Sobel Test Testing of hypothesis mediation can be done with procedure developed by sobel (Ghozali, 2011) and also known as test of sobel. Sobel test is done by testing the strength of indirect effect of X to Y through Z. The Sobel Test formula is the following:

  =√ 2 2 + 2 2 + 2 2

  By Description: sab: The amount of standard error indirect effect a:

  Independent variable path (X) with variable intervening (Z)

  b:

  Variable intervening path (Z) with the dependent variable (Y)

  sa:

  Standard error coefficient a

  sb:

  Standard error coefficient b