A combination multiple comparisons and subset selection procedure to identify treatments that are strictly inferior to the best

This paper considers the problem of identifying which treatments are strictly worse than the best treatment or treatments in a one-way layout, which has many important applications in screening trials for new product development. A procedure is proposed that selects a subset of the treatments containing only treatments that are known to be strictly worse than the best treatment or treatments. In addition, simultaneous confidence intervals are obtained which provide upper bounds on how inferior the treatments are compared with these best treatments. In this way, the new procedure shares the characteristics of both subset selection procedures and multiple comparison procedures. Some tables of critical points are provided for implementing the new procedure, and some examples of its use are given.     

COMPARISONS AND SELECTION OF TWO-PARAMETER EXPONENTIAL POPULATIONS

For two-parameter exponential populations with the same scale parameter (known or unknown) comparisons are made between the location parameters. This is done by constructing confidence intervals, which can then be used for selection procedures. Comparisons are made with a control, and with the (unknown) “best” or “worst” population. Emphasis is laid on finding approximations to the confidence so that calculations are simple and tables are not necessary. (Since we consider unequal sample sizes, tables for exact values would need to be extensive.)     

Inferences on the lognormal mean for complete samples

In many engineering problems it is necessary to draw statistical inferences on the mean of a lognormal distribution based on a complete sample of observations. Statistical demonstration of mean time to repair (MTTR) is one example. Although optimum confidence intervals and hypothesis tests for the lognormal mean have been developed, they are difficult to use, requiring extensive tables and/or a computer. In this paper, simplified conservative methods for calculating confidence intervals or hypothesis tests for the lognormal mean are presented. In this paper, “conservative” refers to confidence intervals (hypothesis tests) whose infimum coverage probability (supremum probability of rejecting the null hypothesis taken over parameter values under the null hypothesis) equals the nominal level. The term “conservative” has obvious implications to confidence intervals (they are “wider” in some sense than their optimum or exact counterparts). Applying the term “conservative” to hypothesis tests should not be confusing if it is remembered that this implies that their equivalent confidence intervals are conservative. No implication of optimality is intended for these conservative procedures. It is emphasized that these are direct statistical inference methods for the lognormal mean, as opposed to the already well-known methods for the parameters of the underlying normal distribution. The method currently employed in MIL-STD-471A for statistical demonstration of MTTR is analyzed and compared to the new method in terms of asymptotic relative efficiency. The new methods are also compared to the optimum methods derived by Land (1971, 1973).     

Pairwise comparisons of matched proportions

Consider the problem of comparing the success rates of c treatments, each of which induce a Bernoulli response. The comparison is to be made on the basis of n matched samples. We present a method for deriving confidence intervals for the pair-wise difference in success rates which has the desirable quality of providing uniformly shorter intervals than the procedure recently proposed by Bhapkar and Somes (1976). Comparisons of the lengths of the respective intervals are provided. Some observations regarding the assumptions required for the use of Cochran’s Q-test (1950) are also made.     

Using biweight m-estimates in the two-sample problem part 1: symmetric populations

We propose replacing the usual Student's-t statistic, which tests for equality of means of two distributions and is used to construct a confidence interval for the difference, by a biweight-“t” statistic. The biweight-“t” is a ratio of the difference of the biweight estimates of location from the two samples to an estimate of the standard error of this difference. Three forms of the denominator are evaluated: weighted variance estimates using both pooled and unpooled scale estimates, and unweighted variance estimates using an unpooled scale estimate. Monte Carlo simulations reveal that resulting confidence intervals are highly efficient on moderate sample sizes, and that nominal levels are nearly attained, even when considering extreme percentage points.     

One and two sample confidence intervals for estimating the mean of skewed populations: an empirical comparative study

In this paper we consider and propose some confidence intervals for estimating the mean or difference of means of skewed populations. We extend the median t interval to the two sample problem. Further, we suggest using the bootstrap to find the critical points for use in the calculation of median t intervals. A simulation study has been made to compare the performance of the intervals and a real life example has been considered to illustrate the application of the methods.     

Multiple comparisons of several populations with more than one control with respectto scale parameter

In this paper, one-sided and two-sided test procedures for comparing several treatments with more than one control with respect to scale parameter are proposed. The proposed test procedures are inverted to obtain the associated simultaneous confidence intervals. The multiple comparisons of test treatments with the best control are also developed. The computation of the critical points, required to implement the proposed procedures, is discussed by taking the normal probability model. Applications of the proposed test procedures to two-parameter exponential probability model are also demonstrated.     

The New Statistics for Better Science: Ask How Much,How Uncertain,and What Else Is Known

Abstract

The “New Statistics” emphasizes effect sizes, confidence intervals, meta-analysis, and the use of Open Science practices. We present three specific ways in which a New Statistics approach can help improve scientific practice: by reducing overconfidence in small samples, by reducing confirmation bias, and by fostering more cautious judgments of consistency. We illustrate these points through consideration of the literature on oxytocin and human trust, a research area that typifies some of the endemic problems that arise with poor statistical practice.     

Relationships between confidence intervals for the ratio of two variances of independent normal distributions

Burk at al (1984) gave a results concerning the comparison of the length of the two different confidence intervals for variance ratio, when the construction of the intervals was based on the principle of “equal tails”11. The purpose of this paper is to be solve the similar problem in case of the principle of “minimal length”.     

Resampling approaches for common intraclass correlation coefficients

In this paper, we present several resampling methods for interval estimation for the common intraclass correlation coefficients. Comparisons are made on the coverage probabilities and average lengths with confidence intervals estimated by using the generalized pivots. Most of the methods proposed in this article produce confidence intervals with better probabilities and shorter average lengths than that produced by using generalized pivots.     

Confidence Intervals for Gini's Diversity Measure and Shannon's Entropy Using Adjusted Proportions

Abstract

Asymptotic confidence intervals are given for two functions of multinomial outcome probabilities: Gini's diversity measure and Shannon's entropy. “Adjusted” proportions are used in all asymptotic mean and variance formulas, along with a possible logarithmic transformation. Exact confidence coefficients are computed in some cases. Monte Carlo simulation is used in other cases to compare actual coverages to nominal ones. Some recommendations are made.     

Adaptive Nonparametric Comparison of Regression Curves

In this article, we develop a new and novel kernel density estimator for a sum of weighted averages from a single population based on utilizing the well defined kernel density estimator in conjunction with classic inversion theory. This idea is further developed for a kernel density estimator for the difference of weighed averages from two independent populations. The resulting estimator is “bootstrap-like” in terms of its properties with respect to the derivation of approximate confidence intervals via a “plug-in” approach. This new approach is distinct from the bootstrap methodology in that it is analytically and computationally feasible to provide an exact estimate of the distribution function through direct calculation. Thus, our approach eliminates the error due to Monte Carlo resampling that arises within the context of simulation based approaches that are oftentimes necessary in order to derive bootstrap-based confidence intervals for statistics involving weighted averages of i.i.d. random variables. We provide several examples and carry forth a simulation study to show that our kernel density estimator performs better than the standard central limit theorem based approximation in term of coverage probability.     

Simultaneous Inferences on the Cumulative Distribution Function of a Normal Distribution

This paper addresses the problem of constructing simultaneous confidence intervals for the cumulative distribution function of a normal distribution at several specified points. The procedure is based upon the observation of a random sample of independent observations from a normal distribution with an unknown mean and variance. A new methodology is proposed for obtaining confidence intervals with a specified overall simultaneous confidence level through the inversion of acceptance sets. Both one-sided and two-sided confidence intervals are considered. Some illustrations of the new method are provided, and comparisons are made with other approaches to the problem.     

Paradoxes and Improvements in Interval Estimation

Cases where confidence sets are empty or include every possible parameter value are an embarrassment to standard theory and difficult to explain to students. To alleviate this problem, and as a convenient way of showing how each parameter value is ranked in view of the data, we propose to plot the confidence set for each possible level, called a confidence curve. Different confidence procedures then lead to different confidence curves. In standard situations involving distributions with a monotone likelihood ratio, we suggest using the confidence curve based on Spjøtvoll's acceptability function. For discrete distributions, this leads to an improvement over usual “exact” confidence intervals.     

Data-randomized confidence intervals for discrete distributions

In practice non-randomized conservative confidence intervals for the parameter of a discrete distribution are used instead of the randomized uniformly most accurate intervals. We suggest in this paper that a part of the data be used as the random mechanism to create “data-randomized” confidence intervals. A thoughtful utilization of the data leads to intervals that are shorter than the usual conservative intervals but avoids the arbitrariness of the randomized uniformly most accurate intervals. Examples are given using the binomial, Poisson, and extended hypergeometric distributions, as well as applications to a metched case-control study and a randomized clinical trial.     

Improved confidence intervals for the exponential mean via tail functions

Abstract

The method of tail functions is applied to confidence estimation of the exponential mean in the presence of prior information. It is shown how the “ordinary” confidence interval can be generalized using a class of tail functions and then engineered for optimality, in the sense of minimizing prior expected length over that class, whilst preserving frequentist coverage. It is also shown how to derive the globally optimal interval, and how to improve on this using tail functions when criteria other than length are taken into consideration. Probabilities of false coverage are reported for some of the intervals under study, and the theory is illustrated by application to confidence estimation of a reliability coefficient based on some survival data.     

Multiple comparisons with the average for exponential location parameters based on doubly censored sample under heteroscedasticity

In this paper, a one-stage multiple comparison procedures with the average for exponential location parameters based on the doubly censored sample under heteroscedasticity is proposed. These intervals can be used to identify a subset which includes all no-worse-than-the-average treatments in an experimental design and to identify better-than-the-average, worse-than-the-average and not-much-different-from-the-average products in agriculture, emerging market, pharmaceutical industries. The critical values are tabulated in a table for practical use. A simulation study on the confidence length and coverage probabilities is done. At last, an example of comparing four drugs in the treatment of leukemia is given to demonstrate the proposed procedures.     

Shrinkage confidence intervals for the normal mean: Using a guess for greater efficiency

If the unknown mean of a univariate population is sufficiently close to the value of an initial guess then an appropriate shrinkage estimator has smaller average squared error than the sample mean. This principle has been known for some time, but it does not appear to have found extension to problems of interval estimation. The author presents valid two‐sided 95% and 99% “shrinkage” confidence intervals for the mean of a normal distribution. These intervals are narrower than the usual interval based on the Student distribution when the population mean lies in such an “effective interval.” A reduction of 20% in the mean width of the interval is possible when the population mean is sufficiently close to the value of the guess. The author also describes a modification to existing shrinkage point estimators of the general univariate mean that enables the effective interval to be enlarged.     

Partitioning with Respect to a Control: Unequal Sample Sizes Case

This article describes a method for partitioning with respect to a control for the situation in which the treatment sample sizes are unequal and also for the situation where the treatment sample sizes are equal except for a few missing values. Calculation of the critical values required for finding confidence limits is discussed and tables are presented for the “almost equal” sample size case. An application of this method to length of stay data for congestive heart failure patients is also provided.     

Exact Confidence Intervals,Based on the Z Statistic,for the Difference Between Two Proportions

We show that the confidence interval version of the extended exact unconditional Z test of Suissa and Shuster (1985) for testing the equality of two binomial proportions is due to general results of Buehler (1957), Sudakov and references cited there (1974), and Harris and Soms (1984). We apply these results to obtain exact unconditional confidence intervals for the difference between two proportions, deriving an explicit solution for the “best” outcome, make some comments on Buehler's (1957) method and give a numerical example. The Appendix contains a listing of the necessary FORTRAN programs.