OBITUARY
Toleti Subba Rao, Granville Tunnicliffe‐Wilson · Journal of Time Series Analysis · 2013
University of Manchester, was an outstanding, influential and highly respected figure in the field of time-series analysis.His book on Spectral Analysis and Time Series, first published in 1982 and reprinted several times, has become a standard reference on the subject.In the preface of the book, Maurice Priestley stated that, for the most part, he 'followed the style of the applied mathematician, but there are certain basic results (such as the spectral representation theorem) which are crucial for a proper understanding of the subject, and by their nature, require a careful and precise presentation'.In his book, one can see his subtle use of real analysis and generalized functions, subjects of which he was very fond, in deriving the results.At the same time, one can see his skill in relating the results to applied problems arising in fields such as physics, engineering and economics.Besides being an outstanding research worker, he was well known to be an excellent teacher.He was one time associate editor of the Journal of the Royal Statistical Society Series B (JRSS B), and from its foundation in 1980 until the end of 2012, he was editor-in-chief of the Journal of Time Series Analysis (JTSA), a leading journal in the field.He was elected as a member of the International Statistical Institute and also a Fellow of the Institute of Mathematical Statistics.Maurice Priestley had a life-long loyalty to Manchester, the city of his birth on 15 March 1933.He attended the 500-year-old renowned Manchester Grammar School and went on to Jesus College, Cambridge, where he was a wrangler in mathematics.At Cambridge, he stayed on to take the Diploma in Mathematical Statistics in the Statistical Laboratory where his interests were influenced by Dennis Lindley and Henry Daniels besides by many others.From 1955 to 1956, he was Scientific Officer at the Royal Aircraft Establishment, gaining practical experience of the relatively new field of spectral analysis and working for a short time with Gwilym Jenkins who was a junior fellow there and who would also go on to make his reputation in this area.They published a joint paper in the JRSS B in 1957, and Maurice decided to study for a PhD.On the advice of George Barnard, he returned to Manchester to work under the guidance of Maurice Bartlett and at the same time took up an assistant lectureship at the University of Manchester.He carried on his work on spectral analysis, which he started with Gwilym Jenkins, and published two very influential papers on spectral estimation (Technometrics, 1962) and the role and choice of bandwidth in the estimation of spectral density function (Applied Statistics, 1965), a topic that received a great deal of attention in the context of probability density estimation and nonparametric regression.His reputation was firmly established with a further two papers in JRSS B based on his PhD topic on inference for mixed spectra -the detection of frequencies in the presence of coloured noise.He published in 1965 a seminal paper (a read paper to the Royal Statistical Society) on evolutionary spectra and non-stationary processes.This paper on nonstationary processes stimulated further research in this area by others and lead to the development of statistical tests for non-stationarity (Priestley and Subba Rao, JRSS B, 1969), to development of time-varying parameter linear autoregressive-moving-average models and to the definition of the weighted likelihood function as a tool for the estimation of time-dependent parameters (Subba Rao, JRSS B, 1970).More recently, Dahlhaus (Annals of Statistics, 1977) and his colleagues extended these ideas further for the analysis of locally stationary processes satisfying autoregressive conditional heteroskedasticity models, a topic that is now receiving considerable attention.During the 1970s, Priestley's department became an exciting centre for time-series research with the appointment, among others, of Tong and Subba Rao who rapidly established their own reputations in nonlinear time-series modelling.One can describe 1970-1980 as a golden decade for rapid developments in nonlinear and non-stationary processes, many of these emanating from the Manchester School under the leadership of Maurice Priestley (Bilinear Models of Subba Rao, JRSS B, 1981; Threshold Models of Tong and Lim, JRSS B, 1980; Amplitude Dependent AR models of Haggan and Ozaki, Biometrika, 1981; State Dependent Models of Priestley, JTSA, 1980).During the same period, Clive Granger and his colleagues at Nottingham were publishing papers in similar and related areas.Together, Maurice Priestley and Clive Granger formed the Manchester-Nottingham Time Series group, which held frequent meetings to discuss new developments that stimulated further research in these areas.These meetings continued until Clive Granger moved in 1974 to the University of California, San Diego, USA.Priestley had been visiting professor at both Princeton and Stanford Universities in the USA in the early 1960s, and through the many contacts he had fostered and his personal standing, he attracted to his department a succession of visitors of international repute in time-series research, including H. Akaike, T. W.