Skip to content

90% of Features Categorical: Can You Still Find High Leverage Points?

In a dataset where 90% of the features are categorical, determining high leverage points remains a valid statistical exercise. These points, which can significantly influence the outcome of a linear regression model, are typically identified using the diagonal elements of the hat matrix. The process involves one-hot encoding the categorical variables to transform them into a format suitable for linear regression analysis. Despite the high proportion of categorical features, the method of using the hat matrix’s diagonal remains applicable. This approach helps in understanding which observations might disproportionately affect the model’s predictions. Therefore, even with a majority of categorical data, computing leverage values is both feasible and recommended for enhancing model accuracy and reliability.

Source: www.reddit.com

Related Videos

Related X Posts

mary @howdymerry · Apr 14
We laughed. We cried. We learned a little about private credit and on chain lending.EP 2.0 WILDCAT FINANCE —————————————————- A quaint conversation and product walkthrough with @functi0nZer0 , founder of @WildcatFi . Wildcat Finance enables

Positions Finance @PositionsFi · Apr 17
Yields are juicy, but they lock you out of liquidity.Introducing Positions Finance — the first-ever on-chain credit protocol — turning your locked assets into composable collateral.Earn, borrow, trade, and leverage—all while still earning yields. /

Kamino @KaminoFinance · Feb 20
1/ JLP Leverage IncreaseUsers can now get up to 4x leverage in JLP Multiply, enabling users to get up to 25% more exposure to JLP

Aaron Harper @AaronHarperCEO · Apr 14
Want to spot a business with HIDDEN LEVERAGE? Here’s my exact processI always analyze both micro AND macro trends before making a move:MICRO TRENDS: Look for businesses that aren’t optimized where YOU can make immediate impactMACRO TRENDS: Identify industries with

Brad Evans, CFA @B_radRiffs · Apr 19
They want to give access to highly levered private companies and levered loans/“private credit” (riskier than High Yield and Junk) to 401k participants?

Plaza Finance @plaza_finance · Feb 28
Programmable derivatives unlock cross-chain composabilityTake your yield (bondETH) or leverage (levETH) position anywhereEarn chain incentives while maintaining your exposureBorrow against or lend out your position for additional leverage or yieldBe free from