In Step 2, linear regression analyses and the decision tree method were performed to answer the research question—How do these associations contribute to the development of news media brand uniqueness?—and test the research hypothesis. Two statistical models were used for all research analyses: linear regression and decision tree techniques. The authors chose one media brand with the highest evaluation scores in attribute strength and uniqueness to build media uniqueness models—delfi.lv. The basis for each model was respondents who ever used the responding news media brand. Accordingly, delfi.lv N = 340. Before modelling, multicollinearity analyses were performed to exclude closely related model factors.
Delfi.lv Brand Uniqueness Model
To test the hypothesis, the authors performed linear regression analyses to see whether and how non-content-related media attributes correlated and developed brand uniqueness. The linear regression-obtained analyses and significance details data are described in
Table A1 in
Appendix A. Linear regression analyses confirmed the hypothesis that non-content-related brand attributes significantly contributed to brand uniqueness. Four media brand attributes correlated with brand uniqueness, significance or
p-value below 0.05. These attributes are one content-related attribute and three non-content-related attributes, thus confirming the research hypothesis that non-content-related brand attributes significantly contribute to brand uniqueness. The attribute with the highest significance (0.011
p-values) to brand uniqueness was
look distinctive and unique. The contribution to uniqueness is high (0.187 unstandardized B) if these associations are strong and favourable. For example, the following attribute playing significance is the
I like journalists, authors content-related attribute (
p-value 0.026, unstandardized B 0.164) and
users can engage in content creation, which is a non-content-related brand attribute (
p-value 0.022, unstandardized B 0.146) that correlates with brand uniqueness.
Table A1 illustrates the linear regression analyses, significance of each media brand attribute, and impact on delfi.lv uniqueness. One of the media brand attributes—the presence of the platforms I use—showed a high significance level (0.009); however, the contribution regarding uniqueness presented with −0.183, leading the authors to believe that this attribute can negatively influence media brand uniqueness.
The first media brand uniqueness model shows delfi.lv attribute correlation to the development of delfi.lv uniqueness association. The decision tree model shows how different brand attributes contribute to news media brand uniqueness. The decision tree method shows how one attribute contributes to the next and forms audience choice accordingly. The decision tree algorithm belongs to the family of algorithms for supervised learning.
In contrast to other supervised learning algorithms, the decision tree approach can be utilised to tackle regression and classification issues. A decision tree aims to develop a training model capable of predicting the class or value of the variable of interest by studying simple choice rules learned from prior data (training data).In machine learning, classification is a two-step process, the learning and prediction steps. The model was developed based on the learning step’s training data (media attributes evaluation). The model predicts the response to the given data in the prediction step. It is also known as a regression tree since the decision or outcome variable is dependent on previous decisions or the type of option involved. One continuous variable decision tree benefit is the ability to predict the outcome based on several variables, as opposed to a single variable, as in a categorical variable decision tree. Predictions are made using decision trees using continuous variables. The tree structure’s simple flowchart structure is one of the fastest methods for identifying significant variables and relationships between two or more variables. The media attributes significance based on weight coefficients. These dependent variables were remodelled in binominal form (yes/no unique) to the valuations 8–10 on the mentioned scale and assigned a “unique” value.
On the other hand, independent variables evaluated 14 brand attributes of particular news media. Similar data accuracy of the model strength coefficients described was applied. The media attributes significance based on weight coefficients. For model strength, the accuracy coefficients were applied—the choice of attributes, based on weighted means, the higher the choice. For decision tree modelling, RapidMiner software was used. The root node is the node that starts the graph. A regular decision tree evaluates the variable that best splits the data [
52]. So, in this research, the root node is the media attribute that is more likely to be a starting point for the audience to form associations of brand uniqueness with attributes of increasing significance in association formation. The root node, or starting point, is the basis for building uniqueness, and the following attributes increase in weight as uniqueness is formed.
None of the attributes are less significant, but each has a unique association-forming effect on significance. For instance, the root node is vital but has the least weight in forming uniqueness and does not determine whether or not uniqueness is generated. This is not the case with other attributes with greater weights. Other attributes increase in significance during the formation process of associations. The overall importance of an attribute in a decision is computed in the following way. Go through all the splits for which the feature was used and measure how much it has reduced the variance or Gini index compared to the parent node. The sum of all importance is scaled to 100 [
52]. This means that each importance can be interpreted as a share of the overall model importance. The model weighs each attribute weight based on given attribute evaluation data in this decision process. In the delfi.lv model, four attributes were selected as more significant to form media brand uniqueness. As seen in
Table 12, four attributes contribute to brand uniqueness.
The root node or attribute—use of attractive special formats, e.g., blogs, podcasts, and videos—is essential but less critical than others for influencing uniqueness directly. If the audience has strong and favourable associations with this attribute, then the strength of the attribute-users can engage in content creation-is significant. If this attribute is strong, favourable, and associated with the audience of delfi.lv, the brand’s uniqueness will be formed. If these associations are not strong and favorable enough, brand uniqueness is not formed, and no other attributes will be able to form uniqueness.
To summarize, strong and favourable attractive special formats and user engagement directly contribute to uniqueness.
Figure 1 illustrates the decision tree model and each attribute’s influence on brand uniqueness. On the other hand, if the audience does not have strong and favourable associations, then delfi.lv use attractive special formats, such as blogs, podcasts, and videos, then distinctive attributes, i.e., unique look, significantly influence uniqueness formation. If the audience associate delfi.lv look as distinctive and unique, the power of likeness of their authors and journalists associations significant. If these associations are strong and favorable, brand uniqueness is formated. Thus, even delfi.lv did not have strong and favorable special interactive format associations, i.e., the strength of a distinctive, unique look and the authors could successfully build uniqueness associations. Authors’ and journalists’ associations are influential in building brand uniqueness if other experience-based associations are not strong enough. Only with the factor that distinctive and unique looks are strong and favourable enough. These associations are essential to forming uniqueness: user engagement in content, the distinctiveness of look and journalists, and authors.
On the other hand, if distinctive, unique look associations are not strong and favorable, unique associations are not built enough, i.e., above 8+. This is because no other associations have enough weight for uniqueness formation. The model shows the importance of experience-based attributes and demonstrates content-related attribute power to form uniqueness, even though some experience-based attributes are not strong enough. Graph 1 illustrates the significance and order of the delfi.lv brand attributes in the audience’s decision-making process when evaluating the delfi.lv brand’s uniqueness.
The audience highly evaluated those attributes; however, others were evaluated higher. Nevertheless, the strength and favorability of these attributes formed delfi.lv uniqueness. The model represents a high accuracy rate—75.74%; the author states that the model predicts how delfi.lv uniqueness was influenced with 75.74% accuracy.
Interestingly, even the attribute users can engage in content was not evaluated by the audience is very high in importance; this attribute plays a significant role in forming delfi.lv brand uniqueness. Both in stated importance by the audience and from both models’ analyses, a distinctive and unique look is an important attribute that contributes to brand uniqueness, even though special formats are not strong enough. Authors’ and journalists’ associations are essential, too, especially if a special attractive format use and distinctiveness of look are not strong and positive enough.
To summarize, the four brand attributes that significantly contribute to delfi.lv brand uniqueness is: associations of special format use, distinctive and unique look, user engagement in content creations, and delfi.lv authors and journalists; all associations increase in importance sequently if some associations are not met in the audience’s minds. The stronger and more favourable of these attributes are built more substantial uniqueness. The analyses confirm non-content-related attribute significance in the development of brand uniqueness. Three of the four attributes significantly contributing to brand uniqueness are non-content-related.