Interaction: Fisher’s Optimal Scores Revisited

  • John C. Gower
  • Sugnet Gardner-Lubbe
  • Niel J. Le RouxEmail author


The normal way of fitting a bilinear model to values of a quantitative variable classified by rows and columns is first to fit the main effects of rows and columns and then fit a biadditive model to the residual table. This gives the ordinary least squares optimal fit, usually visualised by biplots and, among other applications areas, is routinely used in genotype \(\times \) environment trials. When the dependent variable is categorical, Fisher (1938) proposed scoring the categories by maximising the sum-of-squares of the row and column main effects relative to the residual sum-of-squares; then, the usual biadditive analysis again becomes available, although not used by Fisher. In this paper, it is shown that with categorical variables Fisher could have derived scores by maximising the sum-of-squares of the biadditive part of the model relative to the residual sum-of-squares. With Fisher’s data, this would have three advantages: (i) a better fitting model, (ii) a more parsimonious model and (iii) novel and more powerful visualisation techniques for expressing interactions. Other ways of partitioning the contributions from main effects and biadditivity are also discussed. The results show how small changes in model construction can have profound effects on interpretation. Potential applications extend far beyond the data used by Fisher to whenever the categorical nature of the dependent variable offers the freedom to generate quantifications with generalised interaction parameterisations. Supplementary materials accompanying this paper appear on-line.


Biadditive models Biplots Categorical data Modelling Multiplicative interactions Non-additive interactions Optimal scores 

Supplementary material

13253_2017_311_MOESM1_ESM.docx (389 kb)
Supplementary material 1 (docx 389 KB)


  1. Choulakian, V. (1996). Generalized bilinear models. Psychometrika 61, 271–283.Google Scholar
  2. Cox, D. R. (2016). Some pioneers of modern statistical theory: A personal reflection. Biometrika 103, 747–759.Google Scholar
  3. Denis, J-B. and Gower, J. C. (1994). Biadditive models. Biometrics 50, 310–311.Google Scholar
  4. Denis, J-B., and Gower, J. C. (1996). Asymptotic confidence regions for biadditive models: Interpreting genotype-environment interactions. Applied Statistics 45, 479–493.Google Scholar
  5. De Falguerolles, A. and Francis, B. (1992). Algorithmic approaches for fitting bilinear models, in Y. Dodge and J. Whittaker (Eds), COMPSTAT 1992, Physica-Verlag, Heidelberg, 77–82.Google Scholar
  6. De Leeuw, J., Young, F. W., and Takane, Y. (1976). Additive structure in qualitative data: An alternating least squares method with optimal scaling features. Psychometrika 41, 471–503.Google Scholar
  7. Fisher, R. A. (1938). Statistical Methods for Research Workers (7\({\rm th}\) edition). Oliver and Boyd, London.Google Scholar
  8. Fisher, R. A. and Mackenzie, W. A. (1923). Studies in crop variation. II, The manurial response of different potato varieties. Journal of Agricultural Science  13, 311–320.Google Scholar
  9. Gauch, H. G. (1992). Statistical analysis of regional yield trials: AMMI analysis of factorial designs. Elsevier Science Publishers, Amsterdam.Google Scholar
  10. Gifi, A. (1990). Nonlinear multivariate analysis. John Wiley & Sons Ltd., Chichester.zbMATHGoogle Scholar
  11. Gilbert, N. (1963). Non-additive combining abilities. Genetics Research 4, 65–73.Google Scholar
  12. Gollob, H. F. (1968). A statistical model which combines features of factor analytic and analysis of variance techniques, Psychometrika 33, 73–115.Google Scholar
  13. Gower, J. C. (1990). Fisher’s optimal scores and multiple correspondence analysis. Biometrics 46, 947–961.Google Scholar
  14. Gower, J. C., Le Roux, N. J. and Gardner-Lubbe, S. (2015). Biplots: Quantitative data, WIREs, Computational Statistics 7, 42–62.Google Scholar
  15. Gower, J. C., Le Roux, N. J. and Gardner-Lubbe, S. (2016). Biplots: Qualitative data, WIREs, Computational Statistics 8, 82–111.Google Scholar
  16. Gower, J. C., Lubbe, S. and Le Roux, N. J. (2011). Understanding biplots. John Wiley and Sons Ltd., Chichester.CrossRefGoogle Scholar
  17. Guttman, L. (1941). The quantification of a class of attributes: a theory and method of scale construction. In P. Horst (Ed.), The prediction of personal adjustment. New York: SSRC.Google Scholar
  18. Kruskal, J. B. (1965). Analysis of factorial experiments by estimating monotonic transformations of the data. Journal of the Royal Statistical Society, B 27, 251–263.Google Scholar
  19. Martens, H. and Martens, M. (2000). Multivariate analysis of quality: An introduction. John Wiley and Sons Ltd., Chichester.zbMATHGoogle Scholar
  20. McCullagh, P. and Nelder, J. A. (1989), Generalized linear models. Chapman & Hall/CRC, Boca Raton, Florida.CrossRefzbMATHGoogle Scholar
  21. Naes, T., Brockhoff, P.B. and Tomic, O. (2010). Statistics for sensory and consumer science. John Wiley and Sons Ltd., Chichester.CrossRefGoogle Scholar
  22. Van Eeuwijk, F. A. (1993). Practicalities of the use of diagnostic biplots in model choice for two-way tables. In: Proceedings of the 4 \(^{th}\) Conference of the International Federation of Classification Societies, 85–86. Paris: E.N.S.T.Google Scholar
  23. Van Eeuwijk, F. A. (1995). Multiplicative interaction in generalized linear models. Biometrics 51, 1017–1032.Google Scholar

Copyright information

© International Biometric Society 2017

Authors and Affiliations

  1. 1.Department of Mathematics and StatisticsThe Open UniversityMilton KeynesUK
  2. 2.Department of Statistics and Actuarial ScienceStellenbosch UniversityMatielandSouth Africa

Personalised recommendations