Abstract
In this paper, we propose a family of correlation structures for crossover designs with repeated measures for both, Gaussian and non-Gaussian responses using generalized estimating equations (GEE). The structure considers two matrices: one that models between-period correlation and another one that models within-period correlation. The overall correlation matrix, which is used to build the GEE, corresponds to the Kronecker between these matrices. A procedure to estimate the parameters of the correlation matrix is proposed, its statistical properties are studied and a comparison with standard models using a single correlation matrix is carried out. A simulation study showed a superior performance of the proposed structure in terms of the quasi-likelihood criterion, efficiency, and the capacity to explain complex correlation phenomena patterns in longitudinal data from crossover designs.
Similar content being viewed by others
References
Basu S, Santra S (2010) A joint model for incomplete data in crossover trials. J Stat Plan Inference 140(10):2839–2845
Biabani M, Farrell M, Zoghi M, Egan G, Jaberzadeh S (2018) Crossover design in transcranial direct current stimulation studies on motor learning: potential pitfalls and difficulties in interpretation of findings. Rev Neurosci 29(4):463–473
Boik RJ (1991) Scheffés mixed model for multivariate repeated measures: a relative efficiency evaluation. Commun Stat 20(4):1233–1255
Carey VJ (2019) Gee: Generalized Estimation Equation Solver. R package version 4.13-20. https://CRAN.R-project.org/package=gee
Cordeiro GM (2004) On Pearson’s residuals in generalized linear models. Stat Probab Lett 66(3):213–219
Cordeiro GM, McCullagh P (1991) Bias correction in generalized linear models. J R Stat Soc Ser B (Methodological) 53(3):629–643
Cox DR, Snell EJ (1968) A general definition of residuals. J R Stat Soc Ser B 30(2):248–265
Curtin F (2017) Meta-analysis combining parallel and crossover trials using generalised estimating equation method. Res Synth Methods 8(3):312–320
Davis CS (2002) Statistical methods for the analysis of repeated measurements. Springer, San Diego
Diaz FJ, Berg MJ, Krebill R, Welty T, Gidal BE, Alloway R, Privitera M (2013) Random-effects linear modeling and sample size tables for two special crossover designs of average bioequivalence studies: the four-period, two-sequence, two-formulation and six-period, three-sequence, three-formulation designs. Clin Pharmacokinet 52(12):1033–1043
Dubois A, Lavielle M, Gsteiger S, Pigeolet E, Mentré F (2011) Model-based analyses of bioequivalence crossover trials using the stochastic approximation expectation maximisation algorithm. Stat Med 30(21):2582–2600
Forbes AB, Akram M, Pilcher D, Cooper J, Bellomo R (2015) Cluster randomised crossover trials with binary data and unbalanced cluster sizes: application to studies of near-universal interventions in intensive care. Clin Trials 12(1):34–44
Grayling MJ, Mander AP, Wason JM (2018) Blinded and unblinded sample size reestimation in crossover trials balanced for period. Biometr J 60(5):917–933
Hao C, von Rosen D, von Rosen T (2015) Explicit influence analysis in two-treatment balanced crossover models. Math Methods Stat 24(1):16–36
Hardin JW, Hilbe J (2003) Generalized estimating equations. Chapman & Hall, Boca Raton
Harville DA (1997) Matrix algebra from a statistician’s perspective, vol 1. Springer, New York
Hin L-Y, Wang Y-G (2009) Working-correlation-structure identification in generalized estimating equations. Stat Med 28(4):642–658
Hinkelmann K, Kempthorne O (2005) Design and analysis of experiments. Wiley series in probability and mathematical statistics. Applied probability and statistics. Wiley, New York, p 2
Jaime GO (2019) Uso de Un Residuo de Papel Como Suplemento Para Vacas Lecheras. Universidad Nacional de Colombia, Sede Bogota, Tesis de maestría
Jankar J, Mandal A (2021) Optimal crossover designs for generalized linear models: an application to work environment experiment. Stat Appl 19(1):319–336
Jankar J, Mandal A, Yang J (2020) Optimal crossover designs for generalized linear models. J Stat Theory Pract 14(2):1–27
Jones B, Kenward MG (2015) Design and analysis of cross-over trials, 3rd edn. Chapman & Hall/CRC, Boca Raton
Josephy H, Vansteelandt S, Vanderhasselt M-A, Loeys T (2015) Within-subject mediation analysis in ab/ba crossover designs. Int J Biostat 11(1):1–22
Kitchenham B, Madeyski L, Curtin F (2018) Corrections to effect size variances for continuous outcomes of cross-over clinical trials. Stat Med 37(2):320–323
Krzyśko M, Skorzybut M (2009) Discriminant analysis of multivariate repeated measures data with a Kronecker product structured covariance matrices. Stat Papers 50(4):817–835
Lancaster H (1965) The Helmert matrices. Am Math Mntly 72(1):4–12
Leorato S, Mezzetti M (2016) Spatial panel data model with error dependence: a Bayesian separable covariance approach. Bayesian Anal 11(4):1035–1069
Li F, Forbes AB, Turner EL, Preisser JS (2018) Power and sample size requirements for gee analyses of cluster randomized crossover trials. Stat Med 1:1
Li F, Forbes AB, Turner EL, Preisser JS (2019) Power and sample size requirements for gee analyses of cluster randomized crossover trials. Stat Med 38(4):636–649
Liang K-Y, Zeger SL (1986) Longitudinal data analysis using generalized linear models. Biometrika 73(1):13–22
Liu F, Li Q (2016) A Bayesian model for joint analysis of multivariate repeated measures and time to event data in crossover trials. Stat Methods Med Res 25(5):2180–2192
Lui K-J (2015) Test equality between three treatments under an incomplete block crossover design. J Biopharm Stat 25(4):795–811
Madeyski L, Kitchenham B (2018) Effect sizes and their variance for ab/ba crossover design studies. Empirical Softw Eng 23(4):1982–2017
McDaniel LS, Henderson NC, Rathouz PJ (2013) Fast pure R implementation of GEE: application of the Matrix package. R J 5:181–187
Oh HS, Ko S-G, Oh M-S (2003) A Bayesian approach to assessing population bioequivalence in a 2 2 2 crossover design. J Appl Stat 30(8):881–891
Pan W (2001) Akaike’s information criterion in generalized estimating equations. Biometrics 57:120–125
Patterson HD (1951) Change-over trials. Journal of the Royal Statistical Society. Series B (Methodological) 13, 256–271
Pitchforth J, Nelson-White E, van den Helder M, Oosting W (2020) The work environment pilot: an experiment to determine the optimal office design for a technology company. PLoS ONE 15(5):0232943
R Core Team: R: A Language and Environment for Statistical Computing. R Foundation for Statistical Computing, Vienna, Austria (2022). R Foundation for Statistical Computing. https://www.R-project.org/
Ratkowsky D, Alldredge R, Evans MA (1992) Cross-over experiments. Statistics A Series of Textbooks and Monographs, Washington
Rosenkranz GK (2015) Analysis of cross-over studies with missing data. Stat Methods Med Res 24(4):420–433
Roy A, Khattree R (2005) On implementation of a test for Kronecker product covariance structure for multivariate repeated measures data. Stat Methodol 2(4):297–306
Shkedy Z, Molenberghs G, Craenendonck HV, Steckler T, Bijnens L (2005) A hierarchical binomial-Poisson model for the analysis of a crossover design for correlated binary data when the number of trials is dose-dependent. J Biopharm Stat 15(2):225–239
Srivastava MS, von Rosen T, Von Rosen D (2008) Models with a Kronecker product covariance structure: estimation and testing. Math Methods Stat 17(4):357–370
Vegas S, Apa C, Juristo N (2016) Crossover designs in software engineering experiments: benefits and perils. IEEE Trans Softw Eng 42(2):120–135
Wang X, Chinchilli VM (2021) Analysis of crossover designs with nonignorable dropout. Stat Med 40(1):64–84
Zhang H, Yu Q, Feng C, Gunzler D, Wu P, Tu X (2012) A new look at the difference between the gee and the glmm when modeling longitudinal count responses. J Appl Stat 39(9):2067–2079
Acknowledgements
Authors are grateful to M.Sc. George Oneiber Jaime Tenjo and Professor Alvaro Wills from the Department of Animal Sciences of Universidad Nacional de Colombia, Sede Bogotá, for providing the records from the Dairy Cattle experiment, and authors are also grateful to the reviewers for their contributions that improved this document
Author information
Authors and Affiliations
Corresponding author
Additional information
Publisher's Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
Supplementary Information
Below is the link to the electronic supplementary material.
Appendix A
Appendix A
Theorem 2
The estimator of \(\pmb {R}(\pmb {\alpha })\) given by \(\pmb {{\hat{\Psi }}}\otimes \pmb {R}_1(\pmb {{\hat{\alpha }}}_1)\) is asymptotically unbiased and consistent.
Proof
First, asymptotic properties of Pearson’s residuals concerning expectation, variance, and residuals are explored. Define the theoretical Pearson residuals of response of the kth of the ith experimental unit at the jth period as:
and adapting the results of Cox and Snell (1968) it is true that:
where
and \({\mathcal {K}}^{ls}\) is the element at position (l, s) of the inverse of the Fisher information matrix, and according to Cordeiro and McCullagh (1991), the bias of \(\hat{\pmb {\beta }}\) (\(B(\hat{\pmb {\beta }})\)) is given by:
where \(\pmb {1}\) is a vector of appropriate size whose entries are all equal to 1, \(\pmb {D}(z_{ijk})\) is a diagonal matrix with elements on the diagonal given by the variance of the estimated linear predictors, i.e., the diagonal of the matrix
i.e., \(z_{ijk}=Var({\hat{\eta }}_{ijk})\) and \(\pmb {X}\) is the design matrix of the parametric effects described in Equation (3). Now, computing the expected values by taking into account the properties of the exponential family, we obtain:
where \(w_{ijk}\) is the element ijk of the diagonal of the matrix \(\pmb {W}\) defined in Equation (A5). From Equation (A7) and the bias of \(\pmb {\beta }\) given in Equation (A4), it follows that:
where \(\pmb {e}_{ijk}\) is a vector of zeros with a 1’s at the ijkth position. From equations (A8) and (A9), we get:
and therefore, from Eqs. (A13), (A14) and (A2):
where
from Eqs. (A10), (A11) and (A12):
and from equations (A4) and (A12), it follows that:
Therefore, from the results (A16) and (A17), we have that:
where \(\pmb {1}\) is a vector of ones and
By theorem 2 of Liang and Zeger (1986), it is known that the GEE estimators of \(\eta _{ijk}\) is consistent and unbiased, i.e.,
Thus from (A15) and (A18), we find that:
and furthermore, by Sect. 3 of Cordeiro (2004) and equations (A20) and (A21), it follows that:
Let
be a matrix whose first column is \(\frac{\pmb {1}}{\sqrt{n}}\) and the following columns are:
where
then the matrix \(\pmb {\Gamma }\) is a Helmert matrix (Lancaster 1965) and therefore:
If \(r_{ijk}\) is defined as the estimated Pearson residual of the ith experimental unit in the jth period and the kth observation, i.e.,
and the matrix \(\pmb {r}_i\) of residuals of the ith individual, where the first row has the L Pearson residuals defined in Equation (8) corresponding to the first period, and the second row of the L corresponding to the second period and so on until completing a matrix with P rows and L columns, i.e.:
By Equation (A20) and the correlation assumption given in Equation (6), it is true that:
And defining \(\pmb {R}\) as:
and since \(\pmb {\Gamma }\) is orthogonal, then \(\pmb {\Gamma }\otimes \pmb {I}_L\) is also orthogonal. Thus (Srivastava et al. 2008):
and according to Equation (A27), we get:
where \(\pmb {Z}\) is:
with \(\bar{\pmb {r}}\) is the matrix of the average residuals defined in Eq. (A26) for each period, that is,
and
Now by the properties of the Pearson residuals, we have that:
and by the properties given in Equation (A25) and because we assume that the experimental units are independent, that is, \(Corr(r_{ijk}, r_{i'j'k'})=0\), for all \(i\ne i'\) and that Equation (3) is true, then:
furthermore,
By the central limit theorem, we get that:
By equations (A22) and (A30) it follows that:
and by equations (A25), (A31) y (A32) that:
and partitioning \(\pmb {Z}_1\) as follows:
Then, it is obtained that
Similarly, it is found that:
Therefore, since \(\pmb {\Psi }_{jj}=1\), \(\forall j=1,\ldots , P\), \(trace(\pmb {\Psi })=1\) from Equation (A33), we get:
By Theorem 2 in Liang and Zeger (1986), it is known that \(\pmb {R}_1(\hat{\pmb {\alpha }}_1)\) is consistent and unbiased for \(\pmb {R}_1(\pmb { \alpha }_1)\), by equations (A35), (A31) and (A32), we have that
is a consistent and asymptotically unbiased estimator for \(\Psi _{jj'}\), which proves the theorem. \(\square \)
Rights and permissions
Springer Nature or its licensor (e.g. a society or other partner) holds exclusive rights to this article under a publishing agreement with the author(s) or other rightsholder(s); author self-archiving of the accepted manuscript version of this article is solely governed by the terms of such publishing agreement and applicable law.
About this article
Cite this article
Cruz, N.A., Melo, O.O. & Martinez, C.A. A correlation structure for the analysis of Gaussian and non-Gaussian responses in crossover experimental designs with repeated measures. Stat Papers 65, 263–290 (2024). https://doi.org/10.1007/s00362-022-01391-z
Received:
Revised:
Accepted:
Published:
Issue Date:
DOI: https://doi.org/10.1007/s00362-022-01391-z