19/08/2019

# 美国教育学论文代写-不同变量的相关性 Table 2 shows the correlation between different variables considered in the model. The correlation between ROA and ROE is significant at 1% level of significance. They are positively correlated with the correlation coefficient being 0.959. The positive correlation is understandable because as the ROA will increase, ROE will also increase as equity is defined as the difference between assets and liabilities.
The most critical aspect of correlation is its impact on multicollinearity in the model. Variables which are correlated often lead to multicollinearity in the model. Market share and concentration are correlated and the coefficient is statistically significant. The coefficient of correlation is 0.6. Although the correlation is intuitive, this can cause multicollinearity in the model. The correlation between market share and GDP is also statistically significant at -0.577. A negative correlation implies that as the GDP increases, the market share of the banks tend to decline. The market share and deposits are also correlated, albeit negatively.
The market concentration is very strongly correlated with the GDP and Total deposits. Both these correlations are at -0.98 and -0.93. This again points towards strong chances of correlation. The GDP and total deposits are also very strongly correlated with the coefficient at 0.93. These strong correlations can cause multicollinearity in the model. However, it is worth noting here that mere existence of collinearity among the variables does not guarantee multicollinearity in the model. Multicollinearity exists in the model when one or more of the explanatory variables are explaining the other dependent variables in the model. This is checked by the Variance Inflation Factor (VIF). A typical sign of multicollinearity is that the model has a very high explanatory power but most (or all) of the variables are statistically not significant.
However, it is worth noting here that these are correlations only and they do not guarantee causation. To check causality between the variables, it is important to check if the model developed using these variables explain the incidence of dependent variable.