Regression Analysis. A latent-class mixed-effects regression analysis is performed in order to identify sub-groups of patients with distinct trajectory patterns of the psychological variable considered. The analysis is performed using the lcmm package of R. Linear, quadratic and cubic models of the change across time are considered. Models with one to six latent growth classes are fit to the data. Models with different latent processes are also produced. Each model runs several times from different sets of initial values (from a grid of 80 initial values) to avoid convergence to local minima. The number of latent growth classes that best fit the data is assessed by identifying the model with the lowest Akaike information criterion (AIC), Bayesian information criterion (BIC) and sample size–adjusted BIC (SABIC). The average posterior probability of class membership should be above 0.7. Finally, the minimal class size should be at least 5% of the total number of patients.
Regression Analysis. In this part, a regression analysis for teacher-student relationships and students‟ level of emotional intelligence will be demonstrated. This analysis will examine the effect of students‟ age, gender and language of instruction on the relationship between students‟ emotional intelligence and their relationships with teachers. A series of partial correlation analyses was conducted to investigate the effect of students‟ age, gender and language of instruction on the relationship between students‟ emotional intelligence and their relationships with teachers. There was no statistically significant effect of the variables on this relationship.
Regression Analysis. As pointed out in item 3, the basic regression model considers as dependent variable the kind of agreement entered into (“parceria” or partnership=1 and land lease=0) and as independent variable, the income (TALQANO), the number of plants competing in a 50km radius (competitors) and the fact that the lessor is an individual or legal entity (PJ). In Table 3 below there is the result of the regression: Table 3 – Probit Regression In all, there were 1,179 observations. Via the LR Chi-square test, the explanatory variables are jointly important to explain the dependent variable. According to the Prob > Chi-Square, we can reject the hypothesis that all the coefficients are statistically equal to 0 at 1% of significance. The Pseudo R2 demonstrated that about 18% of the variation of the dependent variable can be attributed to explanatory variables. By the classification test, it was possible to identify the percentage of correctness of the model, able to correctly classify 94% of the cases, according to the result presented in Table 4. Table 4 – Classification Test
Regression Analysis. A simple correlation analysis between the RA indexes and a performance measure may indicate, but not definitively establish a causal relationship. In this section we take the analysis a step further, by controlling for common factors influencing the performance measures. As the sample data comprises 13 countries and 9 activities/sectors, this sample can be pooled to increase the degrees of freedom for the statistical analysis and to control for country specific and activity/sector specific influences on the performance measures. This has been done by including ‘dummy variables’ for each sector and country92. We do so because there is reason to believe that most of our performance measures experience common factors in sectors and countries. For instance, some countries might account for a higher labour productivity in all sectors due to a better infrastructure or educational system. 92 For example, the dummy variable for Germany takes the value 1 if an observation is from Germany and 0 otherwise. Country fixed effects filter those differences from the data and enable a better identification of the influence of the reserved activities index. Including sector specific effects controls for common trends of a specific sector across countries. For example, due to the nature of the service provided, a sector (for instance construction or tourism) may require more employees per enterprise (average firm size) than other sectors. The sector specific effects eliminate those aggregation differences and other common trends. Given the technical and complex nature of the general regression equation, the details are presented in Appendix F in the form of supporting information as to how the calculations have been made. The tables later in this sub-section set out the resulting coefficients on the reserved activities index for each sector from the regressions93 of each performance measure. Additionally, we report the standard error as a measure of uncertainty of the coefficient. Significance at a 90%, 95% and 99% confidence-interval is reported by *, ** or *** next to the coefficient, respectively. Inspecting the overall regression results, we see that 9 of 45 coefficients are statistically significant with respect to the confidence levels given above. The explained variation in the data, denoted by R², varies between 85 and 96 per cent, which is due to the country and sector fixed effects. For our first performance measure - average firm size – we find a significant negat...
