Actuarial Notes

Exam PCPA

Property and Casualty Predictive Analytics (PCPA) is a two-part requirement for the ACAS: a 2-hour computer-based exam of 40 questions, followed by a project in which the candidate builds a Generalized Linear Model on one or two data sets provided by the CAS and reports on it in a brief technical report. The exam is offered continuously, year-round, and is the prerequisite for the project. Candidates typically sit it after Exams MAS-I, MAS-II and 5, so it assumes a foundational knowledge of predictive analytics.

Learning Objectives

A. Dealing with Data (35%) Candidates should be able to evaluate the data set and manipulate it so it can be used in a predictive analytics model.

  1. Gather and assess the relevance of information from stakeholders in actuarial analysis.
  2. Import, manipulate, and evaluate data sets using generally available programming languages and software packages (e.g., .csv file).
  3. Evaluate the need for variable transformation and apply appropriate transformations to data.
  4. Identify and appropriately manage outliers and missing data.

Readings: ASOP 23 · ASOP 56 · Gelman and Unwin · Grolemund and Wickham · McKinney

B. Model Diagnostics and Selection (35%) Candidates should be able to create and refine a GLM.

  1. Create and run a GLM.
  2. Evaluate and improve a GLM (e.g., create and interpret diagnostics, conduct cross-validation, incorporate offsets, mitigate multicollinearity issues, avoid underfitting and overfitting) for the data provided and business goals.

Readings: De Jong and Heller · Fannin · Frees, Derrig and Meyers · GLM (Monograph) · Venables and Ripley

C. Model Interpretation and Presentation (30%) Candidates should be able to interpret the findings from a predictive analytics model and present their findings to technical and non-technical audiences.

  1. Create and interpret statistical/tabular and graphical/visual representations of data.
  2. Communicate project technical information, including details on methodologies, modeling decisions, and interpretation of output.
  3. Communicate project findings to non-technical audiences, including the implications on business outcomes or decisions.

Readings: Cairo · Few · Knaflic

Source Material

Source Material

  • ASOP 23 - Data Quality (ASB - 2016)
    • A1–A4
  • ASOP 56 - Modeling (ASB - 2019)
    • A1–A4
  • How Charts Lie (Cairo - 2020)
    • C1–C3
  • Generalized Linear Models for Insurance Rating (Goldburd et al. - 2020)
    • B1–B2 (Chapter 7)
  • Generalized Linear Models for Insurance Data (De Jong and Heller - 2008)
    • B1–B2 (Chapters 5, 6, 8 and related code in Appendix)
  • R for Actuaries and Data Scientists with Applications to Insurance (Fannin - 2020)
    • B1–B2 (Chapters 6, 15, 18 and 19)
  • Show Me the Numbers (Few - 2012)
    • C1–C3
  • Predictive Modeling Applications in Actuarial Science Volume 1 (Frees, Derrig and Meyers - 2014)
    • B1–B2 (Chapter 6)
  • Infovis and Statistical Graphics (Gelman and Unwin - 2012)
    • A1–A4
  • R for Data Science (Grolemund and Wickham - 2017)
    • A1–A4
  • Storytelling with Data (Knaflic - 2015)
    • C1–C3
  • Python for Data Analysis (McKinney - 2022)
    • A1–A4
  • Modern Applied Statistics with S (Venables and Ripley - 2002)
    • B1–B2 (pp. 172–176)