Flirting with Models

The Research Library of Newfound Research

Tag: prediction

Estimating from Historical Data

Estimating from historical data requires many assumptions about similarity. Reducing the number of estimated parameters can control model risk.

Jumpy Models Part II: What we can do with jumps once we find them

If detecting jumps is not hard enough, we have to deal with them afterwards. Models must handle jumps in a way that does not introduce excess whipsaw.

Jumpy Models Part I: Why detecting jumps is important for asset allocation models

Detecting jumps in asset prices is important for robust parameter estimation, especially when estimating volatility.

Why We Don’t Predict: Gartman Says He’s “Never Seen Anything Like This”

Building a model around predictions, may be asking for trouble. As the data that was used to predict changes, the model may fall victim to errors.

Predicting, Forecasting and Black Swans

Black swans, by definition, are unknowns unknowns. When we design models, we incorporate rules to mitigate the risk of model failure regardless of cause.

You are about to leave thinknewfound.com and are being redirected to the website for Newfound Research Funds.