Digital Transformation Payday

APPENDIX FSupplementary Analytics for the Fearless

This appendix serves as the backup for the analysis in the main section of the book and adds additional details, which were deemed too complex for the main flow of the book.

The Orthoforest Causal Estimate Approach

Fortunately, based on the advancements of econometric models, we can build on an orthoforest‐based causal estimate model to run detailed elasticity analysis for DIGITALPROXY, logMARKETCAP and more.

What we are doing in simple words is this:

1. Build a generic causal map (constructed from the dataset by leveraging the Python DoWhy library (Sharma and Kiciman 2020)) based on our chosen Ohlson model variables for logMARKETCAP (see Chapter 3) or their extensions (see the different angles of analysis throughout Part III) including dummy variables to reflect fixed industry and time effects.
2. Generate the corresponding causal estimand (also known as our coefficient in the other models) by once more applying the Python DoWhy library (Sharma and Kiciman 2020).
3. Leverage DoWy's interface to the EconML Python library (Oprescu et al. 2019) to run a fully nonparametric orthoforest for this estimand and calculate a mean estimate.
4. Run so‐called refutation tests to validate the robustness of the model (Sharma and Kiciman 2020), which are described further in the next section of this appendix.
5. Add elasticity analysis for each model variable in relation to DIGITALPROXY on our dependent variable logMARKETCAP by estimating elasticities ...