Chapter 5: Results
- 5 Results
- 5.1 Validity
- 5.2 Reliability
- 5.3 Regression Analysis
5.3 Regression Analysis
Regression analysis can be used to determine a causal relationship between a dependent variable (DV) Y and independent variable(-s) (IV) X. Regression can be used to determine the percent of variance in the dependent variable explained by the independents; to rank the relative importance of independents; to assess interaction effects; and to understand the impact of covariate control variables. The variables have to be measured using an interval or ratio scale. In my web-survey all questions are answered using a Likert scale which is an interval scale.
For examining my research model it is best to split the regression into two separate equations:
- The DV 'eWOM' and IV 'Motivation', 'opportunity' and 'Ability' (see Figure 5.1);
- The IV 'Receiver's Purchase Intentions' and IV 'eWOM' and moderating variable (MV) 'Information Sources' (see Figure 5.2).
The first regression (see Figure 5.1) will be analysed using a simple logistic regression. Logistic regression will be used to predict a lineair and causal relationship between a IV and DV. In this case there are three IVs ('Motivation', 'Opportunity', 'Ability') so there will be three different logistic regression analysis.
Figure 5.1: Equation 1
The second regression (see Figure 5.2) is a different case because of the moderator variable. I will test this model by doing the following. Like with the logistic regression analysis, multiple regression analysis will be done in a similar way between a DV and several IVs. I will use multiple regression to test whether certain IV terms are significant predictors of the DV ('Receiver's Purchase Intentions' in this case). The order of variables is available in Appendix B.
Thus, I will have to compute three different regressions in which these three distinct steps have to be tested. The main effect of eWOM is tested first, the main effect of Information Sources is tested second, and the interaction term is tested third. In the compute box, in SPSS, I have to insert the two main effects and multiply them (e.g., “eWOM * Information Sources”). This will create the new variable ('ewom_info') which is the interaction term.
I also checked for multicollinearity so regression (with unweighted summated scores) and correlation analyses were performed. No multicollinearity problems were encountered since the largest variance inflation (VIF) value was 1,384, which was lower than the commonly suggested cut-off value of 10 (Hair et al., 1998), and the more restricted level of 2.5 (Allison, 1999).
Figure 5.2: Equation 2
5.3.1 Equation 1
The R-squared value is the fraction of the variance in the data that is explained by a regression. In this case the fraction of the explained variance in eWOM by motivation, opportunity and ability is 47,4%. Now I know that the variability of 'Motivation', 'Opportunity' and 'Ability' around the regression line is 1-0.474 times the original variance; in other words I have explained 47,4% of the original variability, and are left with 52,6% residual variability. Ideally, I would like to explain most if not all of the original variability. The R-square value is an indicator of how well the model fits the data (e.g., an R-square close to 1.0 indicates that we have accounted for almost all of the variability with the variables specified in the model). An increase of IVs means an increase of information exchange (eWOM) aswell. Significance of the entire regression model: F = 9,166 / 0,288 = 31,809 , p = 0,000.
| Equation | R | R Square | Adjusted R Square |
|---|---|---|---|
| 1 | 0,69 | 0,47 | 0,46 |
| B | Beta | Sig. | |
| (Constant) | 1,75 | 0,00 | |
| Motivation | 0,43 | 0,58 | 0,00 |
| Opportunity | 0,01 | 0,01 | 0,91 |
| Ability | 0,13 | 0,17 | 0,06 |
The regression coëfficiënt B of the IV 'motivation' has a value of 0,432. The line of the reletionship is therefore positive. Significance of the regression coëfficiënt: motivation p = 0,000. Therefore, based on the above analysis I conclude the following:
H4: The higher the level of the receiver’s motivation to process information, the higher the level of information exchange. [supported]
The regression coëfficiënt B of the IV 'opportunity' has a value of 0,007. The line of the reletionship is therefore positive. An increase of 'opportunity' means a very slight increase of information exchange (i.e., eWOM). The model is insignificant and so the regression formula is not allowed. Therefore, based on the above analysis I conclude the following:
H5: The higher the level of the receiver’s opportunity to process information, the higher the level of information exchange. [not supported]
The regression coëfficiënt B of the IV 'Ability' has a value of 0,128. The line of the reletionship is therefore positive. An increase of 'Ability' means an increase of information exchange (i.e., eWOM) aswell. The Beta shows a score |0,165| which shows that 'Ability' does have influence on 'eWOM'. However, the model is slightly insignificant ( p = 0,058 ). Therefore, based on the above analysis I conclude the following:
H6: The higher the level of the receiver’s ability to process information, the higher the level of information exchange. [not supported]
5.3.2 Equation 2
As shown in Appendix B, I have to do a hierarchical regression (Cohen et al., 1985) in which three distinct steps are stipulated. I will follow the same order of variables as shown in Appendix B.
| Model | R | R Square | Adjusted R Square |
|---|---|---|---|
| 1 | 0,33 | 0,11 | 0,1 |
| 2 | 0,45 | 0,2 | 0,19 |
| 3 | 0,48 | 0,23 | 0,21 |
The adjusted R-squared value explains the fraction of the explained variance in 'Re-ceiver's Purchase Intentions' by eWOM is 11,1%. 'eWOM' does not explain a lot of variability but the most important detail is that the receiver has been influenced. The multiple R-squared value explains the fraction of the explained variance in 'Receiver's Purchase Intentions' by several IV (i.e., 'eWOM' and 'Information Sources') is 20,1%. The combination of these two IV increases the influence on a receiver significant. The multiple R value explains the fraction of the explained variance in 'Receiver's Purchase Intentions' by several IV (i.e., 'eWOM' and 'Information Sources') is 23,1%.
| Model | B | Beta | Sig. | |
|---|---|---|---|---|
| (constant) | 1,72 | 0,00 | ||
| 1 | eWOM | 0,30 | 0,33 | 0,00 |
| (constant) | 1,45 | 0,00 | ||
| eWOM | 1,30 | 0,15 | 0,15 | |
| 2 | Information_Sources | 0,29 | 0,35 | 0,00 |
| (constant) | 3,01 | 0,00 | ||
| eWOM | -0,31 | -0,35 | 0,19 | |
| Information_Sources | -0,27 | -0,33 | 0,35 | |
| 3 | ewom_info | 0,15 | 1,05 | 0,046 |
Looking at table 4 the obtained Beta of the IV 'eWOM' has a value of 0,333. The line of the reletionship is therefore positive. An increase of 'eWOM' means an increase of the re-ceiver's purchase intentions aswell. Therefore, based on the above analysis I conclude the following:
H2: eWOM positively impacts the receiver’s purchase intentions. [supported]
The second step shows that the main effect of 'Information Sources' was entered next. 'eWOM' is still in the equation, and the question is whether the new addition explains significant new variance in the DV. The regression coefficient 'Information Sources' is significant (p = 0,001). The positive beta indicates that higher 'Information Sources' is associated with higher receiver's purchase intentions.
The final step shows that the interaction term significantly added new variance (i.e., yielded a significant p-value). The regression coefficient 'ewom_info' is significant (p = 0,046). The large and positive relationship (Beta score |1,05|) tells one that individuals who reported higher levels of the interaction term and therefore has the most influence on 'Receiver's Purchase Intentions'. Therefore, based on the above analysis I conclude the following:
H1: Information Sources postively influence eWOM's influence on the receiver's purchase intention. [supported]
In order to also analyse the strength of weak and strong ties between sender and receiver, within the construct 'Information Sources', I will do this below.
The multiple R value explains the fraction of the explained variance in 'Receiver's Purchase Intentions' by several IV ('eWOM' and 'Strong Ties') is 13,5%. The multiple R value explains the fraction of the explained variance in 'Receiver's Purchase Intentions' by several IV ('eWOM' and 'Weak Ties') is 15,0%.
| Model | R | R Square | Adjusted R Square |
|---|---|---|---|
| (a) | 0,39 | 0,15 | 0,14 |
| (b) | 0,47 | 0,17 | 0,15 |
- (a) Predictors: (Constant), information_strong, eWOM
- (b) Predictors: (Constant), information_weak, eWOM
The regression coëfficiënt B of the IV 'eWOM' and 'Strong Ties' have a value of, respectively, 0,218 and 0,139. The line of the reletionship is therefore positive. An increase of 'eWOM' and 'Strong Ties' means an increase of the receiver's purchase intentions aswell.
The regression coëfficiënt B of the IV 'eWOM' and 'Weak Ties' have a value of, respectively, 0,179 and 0,185. The line of the reletionship is therefore positive. An increase of 'eWOM' and 'Weak Ties' means an increase of the receiver's purchase intentions aswell.
| Model | B | Beta | Sig. | |
|---|---|---|---|---|
| (constant) | 1,61 | 0,00 | ||
| eWOM | 0,22 | 0,25 | 0,01 | |
| 1 | information_strong | 0,14 | 0,22 | 0,03 |
| 2 | (constant) | 1,54 | 0,00 | |
|---|---|---|---|---|
| eWOM | 0,18 | 0,2 | 0,05 | |
| information_weak | 0,19 | 0,27 | 0,01 |
The IV 'eWOM' has the highest absolute Beta score |0,246| and therefore has the most influence on 'Receiver's Purchase Intentions'. The IV 'Strong Ties' has a somewhat lower Beta score |0,219|. The IV 'Weak Ties' has the highest absolute Beta score |0,269| and therefore has the most influence on 'Receiver's Purchase Intentions'. The IV 'eWOM' has a Beta score |0,201|.