Welcome to Jim's home page. I am a retired Professor in the Psychology Department at McGill University, and a retired Associate
Member of Department of Mathematics and
Statistics. I am now living in
My research interests are:
The analysis of samples of curves, images, or other types of functions. This field is described in our new FDA web site, and in the text Functional Data Analysis, Second Edition co-authored by Bernard Silverman, and published by Springer-Verlag, in 2005.
Applications of FDA that I've been involved in include speech production, handwriting, juggling, psychometrics, human growth, econometrics and weather prediction. See our new book on applications of FDA, Applied Functional Data Analysis also published by Springer-Verlag , in 2002.
Bernard Silverman, and I have also developed a set of functions in R, S-PLUS and Matlab for the analysis of functional data, and these are available at my ftp site . You can also find details in our functional data analysis web site.
Along with these functions we distribute the data analyzed in Functional Data Analysis, and the analyses used to compute the figures and other results in the book. These functions are object-oriented, so that many of the usual displays, summaries, and operations available in these systems can also be applied to functional data objects. This software is continually evolving, so that this site should be checked frequently.
An interesting application of FDA to handwriting, "Functional components of variation in handwriting", has appeared in the March 2000 edition in the Journal of the American Statistical Association. A copy of the paper and the data that were analyzed are available by anonymous ftp from my ftp site .
This is closely related to functional data analysis, and involves techniques for the direct estimation of functions, as opposed to estimating functions by using a model defined by parameters. Specifically, my research topics in this area are:
A widely used smoothing method is cubic spline smoothing. This involves fitting a function to data while penalizing the size of its second derivative. This, in effect, ensures that the fitted function will have limited curvature, and therefore appear smooth to the eye. More generally, if we want to estimate the first derivative, we may decide to fit the data while penalizing the size of the third derivative, and thus have a smooth first derivative. In general, estimation of the mth derivative suggests penalizing the derivative of the fitting function of order m+2. I have developed a module for use with Matlab and SPLUS that allows this. It is, in effect, an extension of the SPLUS function smooth.spline, and is called smooth.Pspline. This function calls a Fortran subroutine, and a C version is also available. The Matlab version is called smooth_Pspline, and is written entirely in Matlab code. These functions are available by anonymous ftp from the my ftp site.
More generally than in P-spline smoothing, we can decide to penalize the size of any linear differential operator L. This allows us to ignore components of variation in the smooth that are in the null space or kernel of L. For example, a P-spline with penalty on the derivative of m effectively ignores variation of polynomial form of degree m-1. An SPLUS module is available by anonymous ftp from my ftp site.
It often happens that we know in advance that the fitting function must be both strictly increasing (or decreasing) and smooth. I have developed a technique for monotone smoothing that is, in effect, a a more general type of smoothing spline. This is described in the paper "Estimating smooth monotone functions" that is given in the references below. R, SPLUS and Matlab functions for monotone smoothing are incorporated into the FDA R, S-PLUS and Matlab modules , and available by anonymous ftp from my ftp site.
This is a program for the analysis data from tests, scales and questionnaires. In particular, It displays the performance of items and options within items, as well as other test diagnostics. TestGraf98 comes with an extensive manual, and is available by anonymous ftp from the my ftp site.
Also on this site is an older MS/DOS version called TestGraf, and a Unix version is available by contacting me directly.
TestGraf98 was last revised on 24 July 2001. In the latest revision, there are help buttons and files associated with most of the displays. The files produced by this version are not compatible with those produced by earlier versions, so that data analyzed before the date displayed for this version should be re-analyzed.
A new feature now in the program is the possibility of forcing right answer option characteristic curves to be monotonic. Monotonicity is a powerful smoothing principle as well as being an assumption that is critical to many theoretical analyses in item response theory.
Two new display variables have been added. Arc length is the length along the one-dimensional path in outcome space defined by the latent trait model. This measure of the trait is manifest, not latent, and is invariant under any permissible transformation of the latent trait. The other display variable is expected rest score, where rest score is the score on the test or scale not using the item being examined. Brian Junker and Klaas Sijtsma have argued persuasively in Applied Psychological Measurement that this gives a better impression of features of the item response function, including its monotonicity.
To accompany the rest score in the Display step, there is now the possibility of using observed rest scores to estimate the option response functions in the Analyze step. When the test or scale has a small number of items, this option is probably preferable since it does not contaminate the independent variable used in smoothing by the influence of the dependent variable. Computation time is slightly increased, but not seriously.
Also a part of functional data analysis, registration is the process of aligning important features or characteristics of curves and images by smooth order-preserving nonlinear transformations, called warping functions, of the argument or domain over which the curve or image is defined. R, S-PLUS and Matlab modules for curve registration are available by anonymous ftp from the my ftp site, and are also incorporated into the FDA S-PLUS and Matlab modules. In the ftp site you find a manuscript with sample registration analyses and a review of the registration problem titled "A Guide to Curve Registration."
Although I'm no longer active in this area, I still distribute an MS/DOS program called Multiscale that has hypothesis testing features, and is available with an extensive manual by anonymous ftp from the my ftp site.
· G., Cao, J., Fussmann, G. and Ramsay, J. O. (2007). Estimating a predator-prey dynamical model with the parameter cascade method. Biometrics to appear.
· G., Cao, J. and Ramsay, J. O. (2007). Parameter cascades and profiling in functional data analysis. Computational Statistics 22, 335-351.
· Ramsay, J. O., Hooker, G., Cao, J. and Campbell, D. (2007). Parameter estimation for differential equations: A generalized smoothing approach (with discussion). Journal of the Royal Statistical Society, Series B. 69, 741-796.
· S+ Functional Data Analysis User's Guide. by Clarkson, D. B., Fraley, C., Gu, C. C. and Ramsay, J. O. Book published by Springer-Verlag, 2005.
· Applied Functional Data Analysis, Second Edition by J. O. Ramsay and B. W. Silverman. Book published by Springer-Verlag, 2002.
· Functional Data Analysis by J. O. Ramsay and B. W. Silverman. Book published by Springer-Verlag, 2005.
· Malfait, N. and Ramsay, J.O. (2003). The historical functional linear model, Canadian Journal of Statistics, 31, 115-128.
· Rossi, N., Wang, X. and Ramsay, J. O. (2002) Nonparametric item response function estimates with the EM algorithm. Journal of the Behavioral and Educational Sciences, 27, 291-317.
· Ramsay, J. O. and Ramsey, J. B. (2001) Functional data analysis of the dynamics of the monthly index of non durable goods production. Journal of Econometrics, 107, 327-344.
· Ramsay, J. O. (2000) Differential equation models for statistical functions. Canadian Journal of Statistics, 28, 225-240.
· Ramsay, J. O. (2000) Penalized regression with model-based penalties. Canadian Journal of Statistics, 28, 241-258.
· Ramsay, J. O. (2000) Functional components of variation in handwriting. Journal of the American Statistical Association, 95, 9-15.
· Kneip, A., Li, X., MacGibbon, B., & Ramsay, J. O. (2000) Curve registration by local regression. Canadian Journal of Statistics, 28, 19-30.
· Ramsay, J. O. & Li, X. (1998) Curve registration. Journal of the Royal Statistical Society, Series B, 60, 351-363.
· Ramsay, J. O. (1998) Estimating smooth monotone functions. Journal of the Royal Statistical Society, Series B, 60, 365-375.
· Ramsay, J. O., Munhall, K. G., Gracco,
V. L. and Ostry, D. J. (1996) Functional data analysis of lip motion. Journal
of the Acoustical Society of
· Ramsay, J. O. (1996) A geometrical approach to item response theory. Behaviormetrika, 23, 3-17.