Showing 1–2 of 2 results for author: Kangro, R
-
Quantifying the Estimation Error of Principal Components
Authors:
Raphael Hauser,
Raul Kangro,
Jüri Lember,
Heinrich Matzinger
Abstract:
Principal component analysis is an important pattern recognition and dimensionality reduction tool in many applications. Principal components are computed as eigenvectors of a maximum likelihood covariance $\widehatΣ$ that approximates a population covariance $Σ$, and these eigenvectors are often used to extract structural information about the variables (or attributes) of the studied population.…
▽ More
Principal component analysis is an important pattern recognition and dimensionality reduction tool in many applications. Principal components are computed as eigenvectors of a maximum likelihood covariance $\widehatΣ$ that approximates a population covariance $Σ$, and these eigenvectors are often used to extract structural information about the variables (or attributes) of the studied population. Since PCA is based on the eigendecomposition of the proxy covariance $\widehatΣ$ rather than the ground-truth $Σ$, it is important to understand the approximation error in each individual eigenvector as a function of the number of available samples. The recent results of Kolchinskii and Lounici yield such bounds. In the present paper we sharpen these bounds and show that eigenvectors can often be reconstructed to a required accuracy from a sample of strictly smaller size order.
△ Less
Submitted 27 October, 2017;
originally announced October 2017.
-
Pricing of Warrants with Stock Price Dependent Threshold Conditions
Authors:
Ander Olvik,
Raul Kangro
Abstract:
Warrants with stock price dependent threshold conditions give the right to buy specially issued stocks, if the performance of the stock price satisfies some requirements. Existence of these derivatives changes the price process of the underlying. We show that in the presence of such warrants one cannot assume that the stock market is arbitrage free and that the stock is tradeable at every time mom…
▽ More
Warrants with stock price dependent threshold conditions give the right to buy specially issued stocks, if the performance of the stock price satisfies some requirements. Existence of these derivatives changes the price process of the underlying. We show that in the presence of such warrants one cannot assume that the stock market is arbitrage free and that the stock is tradeable at every time moment with the same price for buying and selling. This means that the usual methods for deriving fair prices for such warrants cannot be used. We start from a simple model for the firm's value process and discuss some ways to specify a related model for the stock price process in the presence of warrants with threshold conditions. We also discuss how indifference pricing approach can be used for pricing such warrants.
△ Less
Submitted 17 March, 2015;
originally announced March 2015.
