# Autoregressive

An autoregressive correlation structure indicates that two observations taken close in time (or space) within an individual tend to be more closely correlated than two observations taken far apart in the same individual. Formally, p^ = 1 and pjkij / increases in value as the absolute difference between j and k falls. As a specific example, a first-order autoregressive (AR-1) correlation structure specifies that pjk = pli~k] where p is the correlation when | j — k\ = 1.

This is used in balanced data sets. No assumption is made about the relative magnitude of the correlation between any two pairs of observations. Formally, pj} = 1 and Pjku^k] is free to take

All correlation coefficients are fixed by the user rather than being estimated from the data. Formally, pn = 1 and pjk[j ¿ k] can take any value between — 1 and -I-1, but this value is fixed prior to the analysis rather than being estimated from the data.

We must first thank Patricia Cahill RN who supervized every peak flow measurement; without her it would have been impossible to conduct the research upon which our practical example was based. We would also like to thank all of the boys who participated in the study for their time, effort and patience. We thank the statisticians, epidemiologists and clinical scientists who commented on earlier versions of the manuscript. We gratefully acknowledge the teaching and advice provided by the Multilevel Models Project Team at the Institute of Education, London. This research was supported in part by grants from the National Health and Medical Research

1. Armitage, P. and Berry, G. Statistical Methods in Medical Research, 2nd edn, Blackwell Scientific Publications, Oxford, 1987, pp. 70, 82, 86,90,104-106,109,143-150,156,273-295,302, 307, 314-316,

2. Feldman, H. A. 'Families of lines: random effects in linear regression analysis', Journal of Applied

3. Goldstein, H. Multilevel Models in Educational and Social Research, Charles Griffin & Company Ltd,

4. Laird, N. 'Longitudinal studies with continuous responses', Statistical Methods in Medical Research, 1,

5. Neuhaus, J. 'Statistical methods for longitudinal and clustered designs with binary responses', Statistical

6. Zeger, S. L. and Liang, K-Y. 'An overview of methods for the analysis of longitudinal data', Statistics in

7. Liang, K-Y. and Zeger, S. L. 'Longitudinal data analysis using generalized linear models', Biometrika, 73,

8. Zeger, S. L. and Liang, K-Y. 'Longitudinal data analysis for discrete and continuous outcomes',

9. Steimer, J-L., Mallet, A. and Mentre, F. 'Estimating interindividual pharmacokinetic variability', in Rowland, M., Sheiner, L. B. and Steiner, J-L. (eds), Variability in Drug Therapy, Raven, New York, 1985,

10. Goldstein, H. 'Multilevel mixed linear modelling analysis using iterative generalized least squares',

11. Gardner, R. M., Crapo, R. O., Jackson, B.R. and Jensen, R. L. 'Evaluation of accuracy and precision of

12. Miller, M. R., Dickinson, S. A. and Hitchings, D. J. 'The accuracy of portable peak flow meters', Thorax,

13. Miller, M. R. and Pedersen, O. F. 'The peak flow working group of the European Respiratory Society. The characteristics and calibration of devices for recording peak expiratory flow', European Respiratory

14. Clark, N. M., Evans, D. and Mellins, R. B. 'Patient use of peak flow monitoring', American Review of Respiratory Diseases, 145, 722-725 (1992).

15. Harm, D. L., Kotses, H. and Creer, T. L. 'Portable peak-flow meters: intrasubject comparisons', Journal of Asthma, 21, 9-13 (1984).

16. Lloyd, B. W. and Ali, M. H. 'How useful do parents find home peak flow monitoring for children with asthma? British Medical Journal, 305, 1128-1129 (1992).

17. Sly, P., Cahill, P., Willet, K. and Burton, P. 'The accuracy of mini peak flow meters following changes in lung function in asthmatic children', British Medical Journal, 308, 572-574 (1994).

18. Altman, D. G. and Bland, J. M. 'Measurement in medicine: the analysis of method comparison studies', Statistician, 32, 307-317 (1983).

19. Bland, J. M. and Altman, D. G. 'Statistical methods for assessing agreement between two methods of clinical measurement', Lancet, i, 307-310 (1986).

20. Zeger, S. L., Liang K-Y. and Albert, P. S. 'Models for longitudinal data: a generalized estimating equation approach', Biometrics, 44, 1049-1060 (1988).

21. Liang, K.-Y., Zeger, S. L. and Qaqish, B. 'Multivariate regression analysis for categorical data', Journal of the Royal Statistical Society, Series B, 54, 3-40 (1992).

22. Hanfelt, J. 'GEE4 documentation', Technical report, John Hopkins University, 1993.

23. Chambers, J. M. and Hastie, T. J. Statistical Models in S, Wadsworth and Brooks, Pacific Grove, 1992.

24. Splus. User's Manual, Statistical Sciences Inc, Seattle, 1991.

25. Carey, V. J. C Implementation of gee for S (9218/7), Harvard University, Boston, 1992.

26. McCullagh, P. and Nelder, J. A. Generalized Linear models, 2nd edn, Chapman and Hall, Oxford, 1989, pp. 21-47, 404-409, 471.

27. Aitkin, M., Anderson, D., Francis, B. and Hinde, J. Statistical Modelling in GLIM, Clarendon Press, Oxford, 1989, pp. 76, 80, 146.

28. Huber, P. J. 'The behaviour of maximum likelihood estimates under non-standard conditions', Proceedings of the Fifth Berkeley Symposium on Mathematical Statistics and Probability, 1, 221-233 (1967).

29. Breslow, N. E. and Clayton, D. G. 'Approximate inference in generalized linear mixed models', Journal of the American Statistical Association, 88, 9-25 (1993).

30. Rasbash, J. and Woodhouse, G. MLn Command Reference. Version 1.0., Institute of Education of the University of London, London, 1995, pp. 43-45, 80.

31. Prosser, R., Rasbash, J. and Goldstein, H. ML3, Software for Three-level Analysis, Institute of Education of the University of London, London, 1991, pp. 12, 18-20.

32. Healy, M. J. R. Nanostat Users' Guide, London School of Hygiene and Tropical Medicine, London, 1987.

33. Minitab Reference Manual, Minitab Inc, State College of Pennsylvania, 1989.

34. SAS Institute Inc. UNIX release 6.09, SAS Institute Inc, Cary, 1993.

35. Groemping, U. GEE: A SAS Macro for Longitudinal Data Analysis, Universitaet Dortmund, Dortmund, 1994.

36. Goldstein, H. 'Nonlinear multilevel models, with an application to discrete response data', Biometrika, 78, 45-51 (1991).

37. Zeger, S. L. and Karim, M. R. 'Generalized linear models with random effects; a Gibbs sampling approach', Journal of the American Statistical Association, 86, 79-86 (1991).