An asymptotic equivalence of the cross-data and predictive estimators

Abstract

Estimators using multiplicative tuning parameters for maximum likelihood estimators in cross-validation are called cross-data estimators in this paper. Single-sample versions of the cross-data estimators have been called predictive estimators in literatures, which are given by maximizing the expected log-likelihood, where the two-fold expectations are taken over the distributions of future and current data using maximum likelihood estimators based on current data. An asymptotic equivalence of the cross-data and predictive estimators is shown, which guarantees an optimality of the predictive estimator when an unknown population parameter vector is replaced by the sample counterpart. Examples using typical statistical distributions are shown.     

Maximum Likelihood Estimation of the Log-Binomial Model

Maximum likelihood estimation of prevalence ratios using the log-binomial model is problematic when the estimates are on the boundary of the parameter space. When the model is correct, maximum likelihood is often the method of choice. The authors provide a theorem, formulas, and methodology for obtaining maximum likelihood estimators of the log-binomial model and their estimated standard errors when the solution is on the boundary of the parameter space. Examples are given to illustrate the method.     

Inference and prediction for the burr type x model based on records

Based on record values, the maximum likelihood, minimum variance unbiased and Bayes estimators of the one parameter of the Burr type X distribution are computed and compared. The Bayesian and non-Bayesian confidence intervals for this parameter are also presented. A Bayesian prediction interval for the sth future record is obtained in a closed form. Based on simulated record values, numerical computations and comparisons between the different estimators are given     

Computation of the exact likelihood function of an arima process

Acceptance of Arima processes as valuable univariate forecasting mechanisms is increasing. Maximum likelihood estimation of parameters is complicated, and least squares approximations are not always satisfactory. The singular vaiue decomposition is used here to determine numericaily accurate values of the likelihood function for a given set of parameter estimates. Suggestions for efficient computational search procedures of maximum likelihood estimators are made.     

Estimation of the parameters of the gamma distribution by means of a maximal invariant

The use of a scale invariance criterion allows estimation of the shape parameter of the two parameter gamma distribution without estimating the scale parameter. Simulation experiments are used to show that the resulting estimators of both parameters are better than the usual maximum likelihood estimators in terms of both bias and mean square error. Approximately unbiased versions of the maximal invariant based estimators are derived and are shown to be as good as approximately unbiased versions of the usual maximum likelihood estimators     

Exponential Parameter Estimation for Data Sets Containing Left and Right Censored Observations

Lifetime data sets may contain both left and right censored observations. Regression, maximum likelihood, and the best iinear unbiased parameter estimates are given for the exponential distribution, along with a comparison of the estimators.     

Estimating the parameters of a truncated weibull distribution

The problem of nonexistence of the maximum likelihood estimators (m.l.e.) with positive probability is investigated for the truncated Weibull distribution. Similar nonexistence of the m.l.e. is known for some other distributions such as truncated exponential, truncated normal, and one parameter truncated gamma. Modified likelihood estimators, which exist with probability one, are given and compared with the m.l.e.     

SURVIVAL ANALYSIS WITH TIME DEPENDENT FRAILTY USING A LONGITUDINAL MODEL

A method of estimation for generalised mixed models is applied to the estimation of regression parameters in a proportional hazards model with time dependent frailty. A parameter representing change over time is introduced and is modelled in turn into a fixed effect, a normally distributed random effect and a longitudinal effect in which the random component relates to the patient characteristics. Both maximum likelihood and residual maximum likelihood estimators are given.     

Moment properties of estimators for a type 1 extreme-value regression model

A regression model is considered in which the response variable has a type 1 extreme-value distribution for smallest values. Bias approximations for the maximum likelihood estimators are pivm and a bias reduction estimator for the scale parameter is proposed. The small sample moment properties of the maximum likelihood estimators are compared with the properties of the ordinary least squares estimators and the best linear unbiased estimators based on order statistics for grouped data.     

A note on comparison and improvement of estimators based on likelihood

The problem of estimation of a parameter of interest in the presence of a nuisance parameter, which is either location or scale, is considered. Three estimators are taken into account: usual maximum likelihood (ML) estimator, maximum integrated likelihood estimator and the bias-corrected ML estimator. General results on comparison of these estimators w.r.t. the second-order risk based on the mean-squared error are obtained. Possible improvements of basic estimators via the notion of admissibility and methodology given in Ghosh and Sinha [A necessary and sufficient condition for second order admissibility with applications to Berkson's bioassay problem. Ann Stat. 1981;9(6):1334–1338] are considered. In the recent paper by Tanaka et al. [On improved estimation of a gamma shape parameter. Statistics. 2014; doi:10.1080/02331888.2014.915842], this problem was considered for estimating the shape parameter of gamma distribution. Here, we perform more accurate comparison of estimators for this case as well as for some other cases.     

Estimation of the SURE model

The paper presents the essentials of the SURE model and the estimation of its parameters β and ω. Two alternative compact representations of the model are being used. The parameter β is estimated by least squares (LS), generalized least squares (GLS) and maximum likelihood (ML) (under normality). For ω two estimators are being considered, viz an LS-related estimator and a maximum likelihood estimator (under normality). Attention is being given to the study of asymptotic properties of all estimators examined. It turns out that the LS-related and ML estimators of ω follow the same asymptotic (normal) distribution. Efficiency comparisons for the various estimators of β conclude the paper.     

Estimating with percentile grouped and truncated date from a scale parameter distribution

Data which is grouped and truncated is considered. We are given numbers n₁<…<n_k=n and we observe X_ni ),i=1,…k, and the tottal number of observations available (N> n_k is unknown. If the underlying distribution has one unknown parameter θ which enters as a scale parameter, we examine the form of the equations for both conditional, unconditional and modified maximum likelihood estimators of θ and N and examine when these estimators will be finite, and unique. We also develop expressions for asymptotic bias and search for modified estimators which minimize the maximum asymptotic bias. These results are specialized tG the zxponential distribution. Methods of computing the solutions to the likelihood equatims are also discussed.     

A testing approach to estimation

Results from the theory of uniformly most powerful invariant tests are used to develop a new parameter estimation procedure. The procedure is used to derive parameter estimators for several important distributions. Results of simulation studies comparing the performances of the new estimators and maximum likelihood estimators are presented.     

On shrinkage estimation of the exponential location parameter

Under the assumption that the exponential distribution is a reasonable model for a given population, some shrinkage estimators for the location parameter based on type 1 and type II censored samples have been derived. It is shown that these estimators dominate maximum likelihood estimators (MLE's) asymptotically under the mean squared error (MSE) criterion. A Monte Carlo study shows a significant improvement of our estimators over MLE's in terms of MSE for small samples.     

A note on estimation in bernoulli trials with dependence

Finite sample properties of estimators for the parameters of a dependent Bernoulli process are investigated using Monte Carlo techniques. A ratio estimator is proposed for the dependence parameter of the model and is compared to the approximate maximum likelihood estimator given by Klotz. It is shown that both estimators have a downward bias that is extreme in certain cases and that samples well in excess of 200 may be necessary before the asymptotic theory can be applied.     

Modified maximum likelihood and modified moment estimators for the three-parameter weibull distribution

This article is concerned with modifications of both maximum likelihood and moment estimators for parameters of the three-parameter Wei bull distribution. Modifications presented here are basically the same as those previously proposed by the authors (1980, 1981, 1982) in connection with the lognormal and the gamma distributions. Computer programs were prepared for the practical application of these estimators and an illustrative example is included. Results of a simulation study provide insight into the sampling behavior of the new estimators and include comparisons with the traditional moment and maximum likelihood estimators. For some combinations of parameter values, some of the modified estimators considered here enjoy advantages over both moment and maximum likelihood estimators with respect to bias, variance, and/or ease of calculation.     

Comparison of conditional and unconditional confidence intervals for the double exponential distribution

Conditional confidence intervals for the location parameter of the double exponential distribution based on maximum likelihood estimators conditioned on a set of ancillary statistics and the corresponding unconditional confidence intervals based on the maximum likelihood estimators alone are compared in two ways. Monte Carlo techniques are used and the conditional approach appears to give slightly better results although agreement as n becomes larger is noted     

On parameter estimation in Bernoulli trials with dependence

A simple estimator is proposed for the dependence parameter for the Klotz model of Bernoulli trials with Markov dependence and it is compared with the ratio estimator given by Price and the approximate maximum likelihood estimator given by Klotz. The proposed estimator is shown to have considerably smaller bias than the other two estimators with comparable mean squared errors, and has all the large sample optimal properties that the other two estimators have.     

On the estimation of the parameters of the von mises distribution

The usual maximum likelihood estimators of the parameters of the von Mises distribution are shown to perform badly in small samples. In view of this and the fact that these estimators require a large amount of computation, alternative, simpler estimators are proposed. It is shown that these estimators are at least comparable to the traditional estimators and are, in many cases, superior to them. We also apply the procedure of jackknifing to the maximum likelihood estimator of the concentration parameter of the von Mises distribution and compare the properties of the jackknifed estimator with the other estimators considered in this paper.     

Estimation for 3-parameter lognormal distribution with unknown shifted origin

A new reparameterization of a 3-parameter lognormal distribution with unknown shifted origin is presented by using a dimensionless parameter. We avoid, in this article, the application of logarithmic and exponential transformations to a value which has a physical dimension. The distribution function contains two dimensional parameters and one dimensionless parameter. Modified moment estimators and maximum likelihood estimators are presented. The presented modified moment estimators and maximum likelihood estimators are confronted with some actual data.