Abstract
Consider an experiment in which the primary objective is to determine the significance of a treatment effect at a predetermined type I error and statistical power. Assume that the sample size required to maintain these type I error and power will be re-estimated at an interim analysis. A secondary objective is to estimate the treatment effect. Our main finding is that the asymptotic distributions of standardized statistics are random mixtures of distributions, which are non-normal except under certain model choices for sample size re-estimation (SSR). Monte-Carlo simulation studies and an illustrative example highlight the fact that asymptotic distributions of estimators with SSR may differ from the asymptotic distribution of the same estimators without SSR.
Similar content being viewed by others
Change history
12 September 2019
Unfortunately, due to a technical error, the articles published in issues 60:2 and 60:3 received incorrect pagination. Please find here the corrected Tables of Contents. We apologize to the authors of the articles and the readers.
Notes
Other options for consistent variance estimation include bootstrap and cross-validation.
References
Brannath W, Koenig F, Bauer P (2006) Estimation in flexible two stage designs. Stat Med 25(19):3366–3381
Brannath W, Gutjahr G, Bauer P (2012) Probabilistic foundation of confirmatory adaptive designs. J Am Stat Assoc 107(498):824–832
Bretz F, Koenig F, Brannath W, Glimm E, Posch M (2009) Adaptive designs for confirmatory clinical trials. Stat Med 28(8):1181–1217
Broberg P, Miller F (2017) Conditional estimation in two-stage adaptive designs. Biometrics 73(3):895–904
Crowder MJ (1976) Maximum likelihood estimation for dependent observations. J R Stat Soc Ser B 38:45–53
Dispenzieri A, Katzmann JA, Kyle RA, Larson DR, Therneau TM, Colby CL, Clark RJ, Mead GP, Kumar S, Melton LJ (2012) Use of nonclonal serum immunoglobulin free light chains to predict overall survival in the general population. Mayo Clin Proc 87:517–523
Dmitriev Y, Tarassenko P, Ustinov Y (2014) On estimation of linear functional by utilizing a prior guess. In: Dudin A, Nazarov A, Yakupov R, Gortsev A (eds) Information technologies and mathematical modelling. Springer, Cham, pp 82–90
Ferguson T (1996) A course in large sample theory. Routledge, New York
Graf AC, Gutjahr G, Brannath W (2016) Precision of maximum likelihood estimation in adaptive designs. Stat Med 35(6):922–941 sim.6761
Ivanova A, Flournoy N (2001) A birth and death urn for ternary outcomes: stochastic processes applied to urn models. Chapman and Hall/CRC, Boca Raton
Ivanova A, Rosenberger WF, Durham SD, Flournoy N (2000) A birth and death urn for randomized clinical trials: asymptotic methods. Sankhy Ser B 62(1):104–118
Jeganathan P (1982) On the asymptotic theory of estimation when the limit of the log-likelihood ratios is mixed normal. Sankhya: Indian J Stat Ser A (1961-2002) 44(2):173–212
Koenker R (2005) Quantile regression. Econometric Society Monographs. Cambridge University Press, Cambridge
Lane A, Flournoy N (2012) Two-stage adaptive optimal design with fixed first-stage sample size. J Probab Stat 2012:n/a–n/a
Le Cam LM (1960) Locally asymptotically normal families of distributions. Univ Calif Publ Statist 3:37–98
Li G, Shih WJ, Xie T, Lu J (2002) A sample size adjustment procedure for clinical trials based on conditional power. Biostatistics 3(2):277–287
May C, Flournoy N (2009) Asymptotics in response-adaptive designs generated by a two-color, randomly reinforced urn. Ann Stat 37(2):1058–1078
Mutze T, Schmidli H, Friede T (2017) Sample size re-estimation incorporating prior information on a nuisance parameter. Pharm Stat 17(2):126–143 pst.1837
Philippou AN, Roussas G et al (1973) Asymptotic distribution of the likelihood function in the independent not identically distributed case. Ann Stat 1(3):454–471
Proschan MA, Hunsberger SA (1995) Designed extension of studies based on conditional power. Biometrics 51(4):1315–1324
Proschan MA, Lan KKG, Wittes JT (2006) Statistical monitoring of clinical trials: a unified approach. Springer, Berlin
Roussas GG, Bhattacharya D (2011) Revisiting local asymptotic normality (LAN) and passing on to local asymptotic mixed normality (LAMN) and local asymptotic quadratic (LAQ) experiments. In: Advances in directional and linear statistics. Springer, Berlin, pp 253–280
Sugiura N et al (1973) Asymptotic non-null distributions of the likelihood ratio criteria for covariance matrix under local alternatives. Ann Stat 1(4):718–728
Tarima S, Pavlov D (2006) Using auxiliary information in statistical function estimation. ESAIM: Probab Stat 10:11–23
Tarima S, He P, Wang T, Szabo A (2016) Interim sample size recalculation for observational studies. Obs Stud 2(2):65–85
Van der Vaart A (1998) Asymptotic statistics. Cambridge University Press, Cambridge
Wang H, Flournoy N, Kpamegan E (2014) A new bounded log-linear regression model. Metrika 77:695–720
Wang Y, Li G, Shih WJ (2010) Estimation and confidence intervals for two-stage sample-size-flexible design with LSW likelihood approach. Stat Biosci 2(2):180–190
Acknowledgements
We thank Dr. Assaf P. Oron, an unknown referee and the editor for their very valuable feedback on our manuscript.
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.
Appendix
Appendix
Proof (Theorem 1)
The joint asymptotic density (8) is fully determined by \(\left( \widehat{\theta }_{1},\widehat{\eta }_{1}, \widehat{\theta }_{2},\widehat{\eta }_{2}\right) \), and marginally under \(\theta =h/\sqrt{n}\),
By the continuous mapping theorem and the consistency of MLEs, \(\sigma _{\theta }\) evaluated as a function of \(\widehat{\theta }_1\) and \(\widehat{\eta }_1\) is a consistent estimator of \(\sigma _{\theta }\) and using it in place of \(\sigma _{\theta }\), (16) still holds. Similarly, conditionally on N,
Recalling that, with large n and \(N-n\), the full sample MLE is approximately distributed as \((n/N)\widehat{\theta }_1 + (1-n/N)\widehat{\theta }_2\), a large sample approximation follows:
and
It is possible that, for some values of \(\widehat{\theta }_1\) and \(\widehat{\eta }_1\), the total sample size N is not large enough to claim normality; but the probability of this event converges to zero. This follows directly from the equality
The event \(n < N \le n + \sqrt{n}\) includes all possible values of N where the asymptotic normality of V may not hold. The probability of this event
when N / n converges to a stationary random variable with a right-continuous distribution at 1.
Then, for large n,
Equations (18) and (19) are not defined for \(\tau \le 1\) which corresponds to values of \(\widehat{\theta }_1\) and \(\widehat{\eta }_1\) associated with \(N \le n\). When the bound is incorporated the proof is complete.
□
Proof (Corollary 1)
The \(\widehat{{{\text{CV}}}}^2\) is asymptotically distributed as a noncentral chi-squared random variable with 1 d.f. whose distribution is fully determined by \({{\text{CV}}}^2\). Then, asymptotically, under \(H_0\), \(\widehat{{{\text{CV}}}}^2\) is a central chi-squared random variable with 1 d.f. which completes the proof.
□
Rights and permissions
About this article
Cite this article
Tarima, S., Flournoy, N. Asymptotic properties of maximum likelihood estimators with sample size recalculation. Stat Papers 60, 373–394 (2019). https://doi.org/10.1007/s00362-019-01095-x
Received:
Revised:
Published:
Issue Date:
DOI: https://doi.org/10.1007/s00362-019-01095-x
Keywords
- Adaptive designs
- Asymptotic distribution theory
- Interim analysis
- Local alternatives
- Maximum likelihood estimation
- Mixture distributions