8 Causal modeling
In cities throughout the United States, the difficulty of escaping poverty is exacerbated by the difficulty in obtaining social services such as job training, mental health care, financial education classes, legal advice, child care support, and emergency food assistance. They are offered by different agencies in disparate locations with different eligibility requirements. It is difficult for poor individuals to navigate this perplexity and avail services that they are entitled to. To counteract this situation, the (fictional) integrated social service provider ABC Center takes a holistic approach by housing many individual social services in one place and having a centralized staff of social workers guide their clients. To better advise clients on how to advance themselves in various aspects of life, the center’s director and executive staff would like to analyze the data that the center collects on the services availed by clients and the life outcomes they achieved. As problem owners, they do not know what sort of data modeling they should do. Imagine that you are a data scientist collaborating with the ABC Center problem owners to analyze the situation and suggest appropriate problem specifications, understand and prepare the data available, and finally perform modeling. (This chapter covers a large part of the machine learning lifecycle whereas other chapters so far have mostly focused on smaller parts.)