CHAPTER 6: DATA ANALYSIS AND PRSENTATION OF RESULTS
6.9 THE ROADMAP USING AMOS
6.9.2 The Initial Model Fit Measures
The column of means gives the regression weights which are the parameter estimates. Based on the Bayesian Structural Equation Modelling summary above, using PPPs to estimate both sectoral support (mean = 1.349; s.d. = 0.230) and economic support (mean = 1.142; s.d. = 0.213) have the largest weight, thus implying major influence in the model. Using commercialization commitment to estimate public awareness also has a major influence in the model, with a mean of 1.076, s.d. = 0.192. The variables highlighted have insignificant regression weights since their 95% confidence intervals include zero.
On the role played by PPPs to ensure successful R&D–C of research, it is noted that the regression weights are noticeably large, especially the 1.349 value which causes an effect on sectoral support. This implies that there is a big PPPs causal effect, which is confirmed by the qualitative results of the study.
Figure 6.10: The Trace Graph (for the initial model)
The trace left by the simulation of the estimate does not show any unusual behavior. This implies a stable estimate. An autocorrelation graph was created and its tailing off is expected of autocorrelation graphs.
Figure 6.11: The Autocorrelation Graph (for the initial model)
Taking into consideration the issues that emerged from the initial model’s statistics, a revised model was deemed fit. Thus, after excluding the variables which had been deemed insignificant during the initial model analysis, the revised model below emerged and was also exposed to posterior distribution, trace and autocorrelation analyses.
Figure 6.12: The Revised Model (after Removal of Items Excluded from the Initial Model) Source: Developed by the researcher
The polygon of the parameter when using PPPs to estimate Sectoral support has a mean of about 1.25 and is bell shaped, implying that the posterior distribution of this parameter is normally distributed. The trace graph of the estimation of the parameter when using PPPs to estimate sectoral support did not show any unusual behavior during simulation, which implies that the estimate is a stable one. Autocorrelation was also run and the autocorrelation graph of the parameter when using PPPs to estimate sectoral support is exponentially dying out (as illustrated in the figure below) signaling a good fit of the parameter in the model.
Figure 6.13: The Autocorrelation Graph for the Second Phase of the Model
A Bayesian Structural Equation Modelling was done and the table below summarizes the regression weights.
Table 6.29: Bayesian SEM for the Second Stage of the Model
Regression weights Mean S.E. S.D. C.S. Median 95% Lower
bound
95% Upper bound
Skewness Kurtosis Min Max
Market Training<--Commercialization Commitment 0.311 0.001 0.286 1.000 0.309 -0.250 0.880 0.038 0.210 -1.178 1.644
Test Marketing<--Commercialization Commitment 0.932 0.001 0.254 1.000 0.924 0.453 1.453 0.192 0.262 -0.214 2.194
Information Dissemination<-- Commercialization Commitment 0.956 0.001 0.169 1.000 0.949 0.643 1.309 0.268 0.319 0.276 1.922
Feedback Use<--Commercialization Commitment 0.432 0.001 0.208 1.000 0.428 0.031 0.852 0.111 0.276 -0.535 1.650
Technological Infrastructure<--PPPs 0.211 0.001 0.210 1.000 0.210 -0.201 0.625 0.013 0.164 -0.933 1.202
Resource Commitment<--Commercialization Commitment 0.098 0.001 0.226 1.000 0.098 -0.349 0.547 0.013 0.239 -1.075 1.236 Marketing Staff Motivation<--Commercialization Commitment 0.656 0.001 0.192 1.000 0.650 0.295 1.051 0.193 0.258 -0.162 1.640
Accessibility<--Commercialization Commitment 0.819 0.001 0.185 1.000 0.813 0.476 1.204 0.249 0.368 -0.141 1.874
Public Awareness<--Commercialization Commitment 1.078 0.001 0.193 1.000 1.069 0.720 1.484 0.271 0.308 0.333 2.119
Sectoral Support1<--PPPs 1.258 0.001 0.242 1.000 1.247 0.808 1.766 0.280 0.438 0.058 2.771
Location Awareness<--PPPs 0.172 0.001 0.205 1.000 0.173 -0.232 0.572 -0.011 0.150 -0.828 1.138
Intercepts
Sectoral Support 2.546 0.001 0.231 1.000 2.546 2.088 3.002 -0.010 0.189 1.456 3.649
Technological Infrastructure 3.078 0.001 0.154 1.000 3.078 2.773 3.379 -0.011 0.110 2.346 3.787
Accessibility 2.500 0.001 0.171 1.000 2.500 2.163 2.835 -0.006 0.166 1.691 3.307
Location Awareness 3.000 0.001 0.146 1.000 3.000 2.714 3.289 0.013 0.110 2.371 3.743
Resource Commitment 3.269 0.001 0.180 1.000 3.269 2.916 3.625 0.012 0.146 2.413 4.215
Market Training 3.346 0.001 0.233 1.000 3.346 2.883 3.800 -0.020 0.154 2.241 4.452
Test Marketing 3.018 0.001 0.225 1.000 3.018 2.576 3.459 -0.002 0.180 1.937 4.282
Information Dissemination 2.750 0.001 0.166 1.000 2.749 2.422 3.078 0.012 0.156 1.971 3.578
Feedback Use 2.807 0.001 0.175 1.000 2.807 2.461 3.152 -0.007 0.160 1.954 3.681
Marketing Staff Motivation 2.499 0.001 0.169 1.000 2.499 2.165 2.832 0.001 0.151 1.741 3.390
Public Awareness 2.808 0.001 0.189 1.000 2.807 2.436 3.181 0.009 0.150 1.932 3.754
Covariances
PPPS<->Commercialization Commitment 0.550 0.001 0.188 1.000 0.568 0.134 0.861 -0.544 0.309 -0.425 1.230
Variances
e3 1.151 0.002 0.250 1.000 1.119 0.754 1.729 0.867 1.527 0.472 3.150
e4 0.817 0.001 0.203 1.000 0.792 0.494 1.284 0.868 1.625 0.121 2.584
e5 1.044 0.001 0.228 1.000 1.015 0.686 1.573 0.864 1.393 0.415 2.726
e7 1.640 0.002 0.350 1.000 1.594 1.090 2.451 0.887 1.510 0.708 4.229
e9 1.681 0.002 0.404 1.000 1.629 1.040 2.620 0.853 1.456 0.395 4.486
e6 0.669 0.001 0.208 1.000 0.647 0.322 1.135 0.674 1.069 -0.260 2.128
e8 2.632 0.003 0.557 1.000 2.560 1.747 3.927 0.818 1.193 1.167 6.786
e10 0.489 0.001 0.158 1.000 0.473 0.223 0.845 0.667 1.059 -0.039 1.471
e11 1.357 0.001 0.299 1.000 1.317 0.893 2.058 0.913 1.596 0.579 3.545
e12 1.031 0.001 0.239 1.000 1.001 0.656 1.583 0.938 2.011 0.427 3.200
Based on this Bayesian Structural Equation Modelling summary, using PPPs to estimate sectoral support has the largest weight (mean = 1.258; s.d. = 0.242) thus implying it has major influence in this second phase of the model. The variables highlighted in green have insignificant regression weights since their 95% confidence intervals include zero. Thus these variables are dropped in the formulation of the final model. The significance of the regression weights above is shown in the table below.
Table 6.30: The Results indicating Regression Weights: (Group number 1 - Default model)
Estimate S.E. C.R. P Label S.R.W.E.
Market Training <--- Commercialization Commitment .281 .234 1.202 .229 par_1 .181
Test Marketing <--- Commercialization Commitment .834 .204 4.084 *** par_2 .566
Information Dissemination <--- Commercialization Commitment .859 .135 6.374 *** par_3 .802
Feedback Use <--- Commercialization Commitment .384 .171 2.245 .025 par_4 .331
Technological Infrastructure <--- PPPs .212 .182 1.167 .243 par_5 .205
Resource Commitment <--- Commercialization Commitment .089 .183 .483 .629 par_7 .073 Marketing Staff Motivation <--- Commercialization Commitment .586 .157 3.730 *** par_8 .524
Accessibility <--- Commercialization Commitment .736 .149 4.924 *** par_9 .658
Public Awareness <--- Commercialization Commitment .966 .155 6.233 *** par_10 .788
Sectoral Support <--- PPPs 1.154 .205 5.642 *** par_11 .756
Location Awareness <--- PPPs .181 .172 1.052 .293 par_12 .185
Covariances: (Group number 1 - Default model)
Estimate S.E. C.R. P Label
PPPs <--> Commercialization Commitment .574 .162 3.542 *** par_6
Variances: (Group number 1 - Default model)
Estimate S.E. C.R. P Label
PPPs 1.000
Commercialization Commitment 1.000
e1 1.000
e3 1.026 .208 4.936 *** par_24
e4 .708 .163 4.337 *** par_25
e5 .929 .187 4.962 *** par_26
e7 1.458 .288 5.066 *** par_27
e9 1.478 .320 4.625 *** par_28
e6 .568 .165 3.450 *** par_29
e8 2.340 .464 5.039 *** par_30
e10 .411 .124 3.305 *** par_31
e11 1.200 .242 4.956 *** par_32
e12 .907 .192 4.714 *** par_33
The regression weight, 1.154, for PPPs in the prediction of Sectoral Support1 is significantly different from zero at the 0.001 level (two-tailed). Also, the standard errors, the standardized estimates and the
standardized total effects are all below one which is a more desirable scenario. Hence, the model can be used for estimation. However, the regression weights highlighted are not significantly different from zero at the 0.05 level (two-tailed). These have no effect in this model, i.e. they are removed from the final model.
At the 0.05 level of significance, the significant regression weights are for Commercialization Commitment in the prediction of Test Marketing, the regression weight for Commercialization Commitment in the prediction of Information Dissemination, the regression weight for Commercialization Commitment in the prediction of Feedback Use, the regression weight for Commercialization Commitment in the prediction of Marketing Staff Motivation, the regression weight for Commercialization Commitment in the prediction of Accessibility, the regression weight for Commercialization Commitment in the prediction of Public Awareness, and the regression weight for PPPs in the prediction of Sectoral Support. Further, the covariance between PPPs and Commercialization Commitment is significantly different from zero at the 0.001 level (two-tailed).
The paths “Test Marketing<--Commercialization Commitment”, “Information Dissemination<-- Commercialization Commitment”, “Accessibility<-- Commercialization Commitment” have significant regression weights of 0.834, 0.859, and 0.736 respectively. Their magnitudes imply that they are greatly influenced by the unobservable variable “Commercialization Commitment” and therefore in turn impact on the commercialization of TIs. Further, the standardized estimates of the regression weights are less than the threshold of one.
The null hypothesis is to be rejected if the p-value (here denoted by P) is less than 0.05. The null hypothesis here says that “the regression weight is not significant” and the alternative is that “the regression weight is significant”.