/
PRINCIPAL COMPONENT ANALYSIS OF PRINCIPAL COMPONENT ANALYSIS OF

PRINCIPAL COMPONENT ANALYSIS OF - PowerPoint Presentation

tatyana-admore
tatyana-admore . @tatyana-admore
Follow
380 views
Uploaded On 2017-10-22

PRINCIPAL COMPONENT ANALYSIS OF - PPT Presentation

VARIABLE STAR LIGHT CURVES Principal Component Analysis PCA Method developed by Karl Pearson in 1901 Primarily used as a statistical tool in exploratory data analysis Linearly transforms the data matrix into a space where each orthogonal basis vector is ordered in decreasing variance along ID: 598246

fig data pca light data fig light pca ogle akbk principal stars component star curve interpolation column fourier mnras

Share:

Link:

Embed:

Download Presentation from below link

Download Presentation The PPT/PDF document "PRINCIPAL COMPONENT ANALYSIS OF" is the property of its rightful owner. Permission is granted to download and print the materials on this web site for personal, non-commercial use only, and to display it on your personal computer provided you do not modify the materials and that you retain all copyright notices contained in the materials. By downloading content from our website, you accept the terms of this agreement.


Presentation Transcript

Slide1

PRINCIPAL COMPONENT ANALYSIS OF VARIABLE STAR LIGHT CURVES

Principal Component Analysis (PCA)Method developed by Karl Pearson in 1901Primarily used as a statistical tool in exploratory data analysisLinearly transforms the data matrix into a space where each orthogonal basis vector is ordered in decreasing variance along its directionIn this case, we want our data matrix to represent the light curveWhy use PCA to analyze variable stars?Reduces the dimensionality of a problem by providing very good approximations for fewer terms usedCan we link it to a physical property of the star, such as metallicity?Data SetWe have studied stars from two data sets in this project:19 Non-Blazhko RR Lyraes from Kepler1770 LMC Cepheids from OGLE III survey

Ruka Murugan1, Shashi Kanbur21University of Rochester, Rochester, NY2Oswego, Oswego, NY

References:

Kanbur

et.

al., MNRAS 10 October 2002Kanbur et. al., MNRAS 27 June 2004Nemec et. al., MNRAS 01 July 2011Deb & Singh, Astronomy and Astrophysics 08 August 2011

 

METHODS

We have tried several different input matrices to conduct PCA on, and each has yielded interesting results: AkBk Method: each star is a row, reading: Log(P) A1 B1 A2 B2 … Ak Bk Where the A’s and B’s are the Fourier coefficients of the light curve.Interpolation Method: each star is a row, with each column holding the value of the light curve’s amplitude at a given time. We obtain the data by estimating an average from the Fourier fit.Standardization: we employ this is as a scaling for both the AkBk and Interpolation methods, by taking Mij = (Mij – Aj )/Sj, where Aj is the average value of the column, and Sj is the standard deviation of the column..

Fig. 1: Reconstructed light curve comparing a 20th order Fourier fit to 5th order PCA fits.

Light Curve ReconstructionNth order PCA approximations to light curves can be reconstructed by using the formula:Where PC is the principal component value and x is the eigenvector.

 

Fig 2.

Kepler

AkBk

Fig 3.

Kepler

Interpolation

Fig 4. OGLE AkBk unstandardized

Fig 5. OGLE AkBk standardized

Fig 7. OGLE Interp. standardized

Fig 6. OGLE Interp. unstandardized

AnalysisThe next step in the project is to analyze the output of the PCA and determine what data can be used reliably to create an empirical formula. This will involve using further statistical analysis of our principal component values to find correlation between PC, period, magnitude and metallicity. We must also work out a reliable way to scale the data, see how principal components vary in individual stars from cycle to cycle, and determine how sensitive PC values are to small changes in the dataset, such as removing one or several stars from the pool.

NSF Office of International Science and Engineering award number 1065093