4.1.2. Insights for Correlation Analysis Results

In the correlation analysis (see Section 3.2), the main results in Table 2 and Figure 3 reveal that:


Beyond the correlations, additional information is provided. Table 3 includes eight sub tables with correlations produced when variable *X* (of the variable pair (*X*, *Y*)) is converted using the eight transformation methods described in Section 2.3.4. These sub tables confirm the main results in Figure 3 and Table 2 (when no data variable is transformed), and also help identify the optimal transforms to be used for each variable (this is further discussed in Section 4.1.4). In this regard, such information is valuable.

#### 4.1.3. Insights from the Results of the Cosine-Similarity Analysis

By treating the data variables as vectors, the results obtained from the cosine-similarity analysis essentially confirm the correlations identified between pairs of variables. With few exceptions, a higher Cos-Sim index (between 0 and 1) in Table 4 indicates it has also had a relatively high P-Co-Co in Table 2 (between −1 and 1); this result can also be observed by comparing Figures 3 and 4. As such, the main results of the correlation analysis are further confirmed (and 'double checked', in addition to the confirmations provided in Section 4.1.2 by the P-Co-Cos recalculated after different variable transforms).

Another insight of this research relates to methodology and theory. This study offers new support for the claim that a cosine-similarity analysis can be used as a supplement to the traditional correlation analysis.

#### 4.1.4. Insights from the Established SLR Models

In this study, after successfully estimating the model parameters used in the datasets, 504 SLR models are established. The major insights gained from these analyses are summarised as follows.
