Functional Projection Pursuit [1 citations — 0 self]
|
Abstract:
This article describes the adaption of exploratory projection pursuit for use with functional data. The aim is to find "interesting " projections of functional data: e.g. to separate curves into meaningful clusters. Functional data are projected onto low-dimensional subspaces determined by a projection function using a suitable inner product. Such a projection is rapidly computed by representing data and projection function in terms of a suitable orthogonal basis. Both Fourier and wavelet bases are considered and their advantages and disadvantages outlined. The concepts of interpretable projection solutions, centring and sphering are also discussed. Two examples are presented: one simulated and one real concerning the core temperature of a group of infants that sleep with and without their mothers. The aim for the real data is to find interesting projections that may separate the two groups. 1
Citations
| 168 | Exploratory Projection Pursuit – Friedman - 1987 |
| 167 | Functional Data Analysis – Ramsay, Silverman - 1997 |
| 85 | Density Estimation – Silverman - 1996 |
| 55 | Projection pursuit (with discussion – Huber - 1985 |
| 49 | Estimating the Mean and Covariance Structure Nonparametrically When the Data Are Curves – Rice, Silverman - 1991 |
| 31 | What is projection pursuit? (With discussion – Jones, Sibson - 1987 |
| 20 | Canonical correlation analysis when the data are curves – Leurgans, Moyeed, et al. - 1993 |
| 8 | Smoothed Functional Principal Components Analysis by Choice of Norm – Silverman - 1996 |
| 5 | Principal differential analysis: Data reduction by differential operators – Ramsay - 1996 |
| 1 | Interpretable exploratory projection pursuit. Pages 470--474 of – Morton - 1990 |
