When estimating in-game purchases from side-quest engagement, statistical modeling provides a mathematical framework for describing patterns in observed data. Its assumptions determine whether the result supports association alone or stronger causal conclusions. section 1.1 section 10.1.2
A data-generating process produces recorded outcomes. A model simplifies that process mathematically, while a statistical model tunes parameters to align with patterns in the data. section 2.3
Models differ according to the assumptions they place on the data-generating process:
- Associational models use statistical, non-causal assumptions, such as a linear regression or neural network for P(I|E). section 10.1.2
- Interventional models add assumptions represented by a causal directed acyclic graph, enabling conclusions about interventions. section 10.1.2
- Counterfactual models add assumptions about mechanisms, such as how variables affect one another or whether a relationship is naturally linear. section 10.1.2
Engagement and purchases
A model mapping side-quest engagement E to in-game purchases I can predict statistical patterns without claiming that changing E causes I. Causal assumptions move the analysis beyond association. section 10.1.2
