Confidence interval estimation of the common mean of several gamma populations.

Gamma distributions are widely used in applied fields due to its flexibility of accommodating right-skewed data. Although inference methods for a single gamma mean have been well studied, research on the common mean of several gamma populations are sparse. This paper addresses the problem of confidence interval estimation of the common mean of several gamma populations using the concept of generalized inference and the method of variance estimates recovery (MOVER). Simulation studies demonstrate that several proposed approaches can provide confidence intervals with satisfying coverage probabilities even at small sample sizes. The proposed methods are illustrated using two examples.

Introduction

Due to its flexibility of accommodating right-skewed data, the standard two-parameter Gamma distribution has been widely used in many applied fields such as meteorology, reliability, medical science, engineering and quality control [1-4]. Under many circumstances, the research interest lies in making inference about the mean. There exit abundant research regarding making inference about gamma mean(s). For example, Fraser et al. [5] investigated inference methods for gamma mean based on asymptotic approximation, and Krishnamoorthy and León-Novelo [6] investigated small sample inference for gamma parameters for one-sample and two-sample problems. Recently, several fiducial methods [7-9] constructed approximate generalized pivotal quantities for a single gamma mean in different ways. Wang et al. [10] extended a fiducial approach [7] for a single gamma mean to construct a fiducial confidence interval for the difference between two independent gamma means.

There also exist some research on testing equality of several gamma means. For example, Chang et al. [11] proposed a parametric bootstrap method for comparing several gamma means, and Krishnamoorthy et al. [12] presented likelihood ratio test for comparing several gamma distributions. When testing equality of several gamma means concludes the null hypothesis (i.e. all the means are equal) can not be rejected, naturally, making inference about the common mean is of interest. Despite the fact that inference procedures about the common gamma means are of practical and theoretical importance, there has not yet been a well-developed approach for this purpose at small sample sizes except some traditional large sample methods. Therefore, the goal of this paper is to present accurate small sample inference methods for confidence interval estimation for the common gamma mean derived from several independent samples.

The rest of this paper is organized as follows. We will first present preliminaries including notations and existing methods for confidence interval estimation of single gamma mean. Then we will propose several methods for constructing confidence intervals for common gamma mean. Simulation results are presented to evaluate the performance of the proposed methods and examples are analyzed using the proposed methods. Finally, summary and discussion are given.

Preliminaries

The setting

Consider K independent gamma populations. Let be a random sample from the ith gamma population as Y ∼ gamma(α, β) where α is shape parameter and β is rate parameter; i.e. the corresponding probability density function for Y is for y > 0, α, β > 0. Let μ denote the population mean for ith sample. Then μ = α/β for i = 1, 2, …, K. We assume that μ1 = μ2 = … = μ and let μ denote the common mean. The goal of this paper is to present procedures for confidence interval estimation of μ at small to medium sample sizes.

Let and stand for the maximum likelihood estimates for α and β, respectively. The maximum likelihood estimate of μ is where the large sample variance for is [5] and its estimate is

The common gamma mean can be estimated as a pooled estimate of sample means defined as Using standard large sample theory, we have asymptotically. Hence, a simple large sample solution for confidence interval estimation for common μ is

Of course, we also can obtain a large sample confidence interval using standard maximum likelihood theory. However, these large sample solutions do not have good performance at small sample sizes. Hence in this paper, we will present some procedures with satisfactory performance.

Existing methods for confidence interval estimation for single gamma mean

In the following, we will review several existing methods for confidence interval estimation for single gamma mean. These methods are known to have reasonable performance at small to medium sample sizes, and will be used in the following to present our new procedures for confidence interval estimation for common gamma mean.

Let Y1, Y2, …, Y be a random sample from a gamma population gamma(α, β). The population mean μ = α/β. Let and denote the arithmetic mean and geometric mean, respectively. The maximum likelihood estimate of μ is where and are the maximum likelihood estimates for α and β, respectively.

Methods based on generalized inference

The generalized variables and generalized pivots were introduced by Tsui and Weerahandi [13] and Weerahandi [14]. More details can be found in the book by Weerahandi [15]. The concepts of generalized pivotal quantity and generalized confidence interval have been successfully applied to a variety of practical problems where standard exact solutions do not exist and it has been shown that generalized inference method generally have good performance, even at small sample sizes; see e.g. [16-21]. Recently, Hannig et al. [22] demonstrated generalized confidence intervals coincide with fiducial confidence intervals. In the following, we review three existing methods for constructing generalized pivotal quantity for single gamma mean.

Krishnamoorthy and Wang’s method: [7, 23] This method is based on the fact that (j = 1, 2, …, n) follows N(μ, σ2) approximately for gamma distribution. Let and be the observed sample mean and sample variance based on the transformed data . The generalized pivotal quantities for μ and σ2 can be obtained as [24]: where Z ∼ N(0, 1), , and Z and U are independent. Furthermore, the generalized pivotal quantities for for α and β can be written as: Chen and Ye’s method: [8, 25] Note that approximately, where ν = 2E2(V1)/var(V1), c = E(V1)/ν, E(V1) = 2nα(ψ(nα) − ψ(α) − log(n)) and var(V1) = 4n2 α2(ψ′(α)/n − ψ′(α)) with ψ and ψ′ being the digamma and trigamma functions respectively. Then and can be obtained by substituting α with its point estimate . An approximate generalized pivotal quantity (GPQ) for α is: where , and are observed values of and . Furthermore, as , a GPQ for β can be constructed as: where .

Wang and Wu’s method [9]: This method is based on Cornish-Fisher approximation. Let and F(.) be the c.d.f. of T. Note that U = F(T)∼U(0, 1). Using the Cornish-Fisher expansion, the Uth percentile of T can be approximated by g1(α) + [g2(α)]1/2 Q(α, U), where g(α) is the ith cumulant of T and Q(α, U) is a function of g(α)’s. Detailed formula can be found in [9]. Let t denote the observed value of T. Solving t = g1(α) + [g2(α)]1/2Q(α, U) for α, we obtain the approximate R. The GPQ for β can be obtained similarly as in (6): where . This method improves Chen and Ye’s method and can work well even when the shape parameter α is small.

The three aforementioned methods for generating R and R lead to three generalized pivots of a single gamma mean: Via simulation, we can obtain an array of R’s and the estimated confidence interval for μ is (R(α/2), R(1 − α/2)) where R(α) is the 100αth percentile of R’s.

A parametric bootstrap method [6]

Krishnamoorthy and León-Novelo [6] presented a method based on parametric bootstrapping for confidence interval estimation using the following pivotal quantity: where and are based on a bootstrap sample from distribution. A two-sided 100(1 − p)% confidence interval (l, u) for μ is: where Q as the 100pth percentile of Q defined in (9).

The proposed methods for confidence interval estimation of common gamma mean

The methods based on generalized inference

For the ith (i = 1, 2, …, K) sample, we can obtain the generalized pivotal quantities using one of the three methods reviewed above (i.e. Krishnamoorthy and Wang’s method [7, 23] Chen and Ye’s method [8, 25], and Wang and Wu’s method [9]). Replacing μ and α with and in (1), the generalized pivotal quantity for can be written as

The generalized pivotal quantity we propose for the common gamma mean μ is a weighted average of the generalized pivot ’s based on K individual samples, i.e. where .

It is easy to see that R satisfies the two conditions to be an approximate bona fide generalized pivotal quantity: 1) the distributions of R is independent of any unknown parameters; and 2) the observed value of R equals to the common gamma μ approximately. This way of constructing generalized pivots for common mean has been widely used in literature. For example, Krishnamoorthy and Lu [17] studied inferences on the common mean of several normal populations based on the generalized variable method; and Tian and Wu [26] studied common mean of several log-normal populations.

Computing algorithms

Consider a given data set Y’s (i = 1, 2, …, K, j = 1, 2, …, n) where the ith sample is from gamma(α, β). We assume μ = μ for all i = 1, 2, …, K. The generalize confidence intervals for the common mean μ can be computed by the following steps:

Using one of the three methods presented above, generate and , then calculate generalized pivot for μ following (12) for i = 1, 2, …, K.

Repeat steps 1, generate and and calculate . Using and , calculate following (11) for i = 1, 2, …, K.

Using obtained in step 1 and in step 2 for i = 1, 2, …, K, calculate the generalized pivot of the common mean R from (12).

Repeat Steps 1-3 a total B (B = 2000) times and obtain an array of R’s.

Rank this array of R’s from small to large.

The 100αth percentile of R’s, R(α), is an estimate of the lower bound of the one-sided 100(1 − α)% confidence interval and (R(α/2), R(1 − α/2)) is a two-sided 100(1 − α)% confidence interval.

Remark 2.1: In computing algorithm, we used different sets of random variables for and . Our simulation shows that the generalized pivotal quantity based on the same set of random variables for and produces confidence intervals which are too liberal. Similar conclusions have been stated in [17, 26].

We refer these three methods based on the generalized pivots of common gamma mean as GV, GV, GV, corresponding to the methods used for confidence interval estimation of a single gamma mean, i.e. Krishnamoorthy and Wang’s method [7, 23] Chen and Ye’s method [8, 25], and Wang and Wu’s method [9].

The MOVER-type methods

The method of variance estimates recovery (MOVER) is a useful technique for obtaining a closed-form approximate confidence interval for a linear combination of parameters based on the confidence intervals of the individual parameters [27, 28]. In this section, using the methods for estimating confidence intervals for a single gamma mean reviewed above, the MOVER method is applied for confidence interval estimation of the common gamma mean.

Let l and u be the lower and upper limits of an approximate two-sided 100(1 − p)% confidence interval (l, u) for the gamma mean based only on ith sample. A MOVER 100(1 − p)% confidence interval (L, U) of the common gamma mean is given by [27, 28]: where , is defined in (2), and and are the maximum likelihood estimates for α and μ, respectively.

For calculating confidence intervals (l, u) for the single gamma mean μ (i = 1, …, K), we will use the three generalized inference methods (i.e. Krishnamoorthy and Wang’s method [7, 23], Chen and Ye’s method [8, 25], and Wang and Wu’s method [9]) as well as the parametric bootstrap method by Krishnamoorthy and León-Novelo [6] reviewed above. Each method provides an approximate confidence interval (l, u) for the ith single gamma mean μ (i = 1, 2, …K).

Substituting (l, u) in (13), we obtain confidence interval estimation of common mean μ. We refer these MOVER-type methods as MOVER, MOVER, MOVER, MOVER corresponding to the methods used for single gamma mean, i.e. Krishnamoorthy and Wang’s method [7, 23] Chen and Ye’s method [8, 25], Wang and Wu’s method [9], and the parametric bootstrap method by Krishnamoorthy and León-Novelo’s method [6], respectively.

Simulation studies

In previous section, we presented several methods for confidence interval estimation of common gamma mean: three methods based on the generalized pivots (i.e. GV, GV, GV); and four MOVER-type methods (i.e. MOVER, MOVER, MOVER, MOVER).

Simulation studies are carried out to evaluate the performances of proposed methods in terms of coverage probabilities and average lengths of proposed confidence intervals. The number of samples K is set as 2 and 5, and a variety of sample sizes from small (5) to large (50) including balanced and unbalanced settings are used. The parameter settings are as follows: 1) common mean μ is set as 1 and 5; 2) shape parameter for each sample varies from 0.5 or 1 to 5 or 10, and the differences among K shape parameters varies from small to large. For each parameter setting, 2,000 random samples are generated. For the confidence interval based on generalized pivots (i.e. GV, GV, GV), MOVER, MOVER, MOVER), 2000 values of generalized pivots are obtained for each random sample. For the confidence interval based on parametric bootstrapping (MOVER), 2000 bootstrap samples are generated for each random sample. The performances of each method is assessed by coverage probability and average lengths of proposed confidence intervals. The simulation results are presented in Tables 1 and 2.

Table 1

Coverage probabilities (CP) and length of confidence interval (CI) of proposed 95% confidence intervals for the common gamma mean (2000 simulations) with two independent samples (K = 2).

(α1, α2)	Sizes*	GVK		GVC		GVW		MOVERK		MOVERC		MOVERW		MOVERboot
		CP	CI	CP	CI	CP	CI	CP	CI	CP	CI	CP	CI	CP	CI
μ = 1
(0.5,1)	I	0.947	3.798	0.977	31.948	0.962	9.138	0.910	2.927	0.950	28.028	0.927	8.921	0.903	2.670
	II	0.952	2.131	0.976	10.748	0.966	4.693	0.917	2.411	0.952	35.248	0.941	9.745	0.917	2.443
	III	0.955	1.458	0.969	1.896	0.964	1.722	0.916	1.260	0.940	1.628	0.936	1.496	0.927	1.337
	IV	0.947	0.827	0.961	0.933	0.959	0.909	0.921	0.756	0.941	0.847	0.936	0.828	0.935	0.813
	V	0.941	0.465	0.954	0.505	0.951	0.499	0.922	0.446	0.942	0.482	0.939	0.479	0.943	0.477
(1,2)	I	0.965	2.274	0.974	4.504	0.966	2.656	0.927	1.804	0.942	3.353	0.930	2.126	0.917	1.441
	II	0.957	1.319	0.969	2.027	0.960	1.483	0.931	1.464	0.948	2.957	0.936	1.810	0.922	1.202
	III	0.958	0.971	0.966	1.050	0.960	0.992	0.926	0.860	0.933	0.922	0.931	0.879	0.924	0.839
	IV	0.954	0.584	0.960	0.605	0.959	0.593	0.934	0.545	0.945	0.562	0.941	0.554	0.941	0.549
	V	0.945	0.333	0.948	0.341	0.948	0.338	0.933	0.323	0.940	0.330	0.939	0.329	0.938	0.328
(1,10)	I	0.959	1.035	0.964	1.532	0.960	1.108	0.933	0.825	0.946	1.326	0.932	0.927	0.911	0.725
	II	0.967	0.579	0.971	0.798	0.969	0.629	0.948	0.643	0.954	1.187	0.946	0.771	0.934	0.551
	III	0.957	0.478	0.962	0.494	0.961	0.481	0.941	0.434	0.945	0.445	0.941	0.436	0.940	0.427
	IV	0.956	0.293	0.959	0.296	0.959	0.294	0.947	0.281	0.946	0.283	0.947	0.282	0.947	0.281
	V	0.948	0.172	0.951	0.172	0.950	0.172	0.946	0.169	0.947	0.170	0.946	0.170	0.946	0.170
(2,10)	I	0.960	0.891	0.963	1.023	0.953	0.862	0.931	0.727	0.933	0.812	0.927	0.709	0.917	0.647
	II	0.954	0.526	0.957	0.576	0.953	0.525	0.933	0.562	0.940	0.653	0.934	0.560	0.921	0.491
	III	0.964	0.446	0.964	0.453	0.961	0.444	0.943	0.408	0.939	0.411	0.943	0.405	0.935	0.400
	IV	0.949	0.279	0.949	0.280	0.949	0.278	0.940	0.267	0.938	0.268	0.939	0.267	0.935	0.266
	V	0.952	0.166	0.953	0.167	0.953	0.166	0.950	0.164	0.947	0.164	0.950	0.164	0.949	0.164
(5,10)	I	0.964	0.732	0.969	0.762	0.961	0.690	0.932	0.612	0.931	0.626	0.925	0.586	0.920	0.561
	II	0.960	0.478	0.963	0.489	0.960	0.467	0.944	0.497	0.946	0.510	0.936	0.480	0.935	0.456
	III	0.956	0.390	0.956	0.392	0.957	0.387	0.934	0.358	0.936	0.358	0.934	0.355	0.936	0.352
	IV	0.945	0.246	0.949	0.246	0.945	0.245	0.929	0.235	0.931	0.235	0.927	0.235	0.932	0.234
	V	0.957	0.148	0.955	0.148	0.957	0.148	0.949	0.145	0.949	0.145	0.949	0.145	0.948	0.145
μ = 5
(0.5,1)	I	0.956	18.934	0.977	152.736	0.965	46.391	0.923	14.473	0.953	163.612	0.936	46.134	0.907	13.224
	II	0.958	10.803	0.977	49.275	0.969	22.734	0.924	12.269	0.952	184.086	0.936	46.080	0.924	12.116
	III	0.956	7.174	0.974	9.363	0.966	8.501	0.917	6.177	0.941	8.038	0.931	7.393	0.927	6.588
	IV	0.932	4.122	0.950	4.651	0.948	4.523	0.906	3.782	0.928	4.226	0.925	4.133	0.923	4.059
	V	0.944	2.304	0.963	2.498	0.961	2.473	0.928	2.210	0.950	2.385	0.948	2.369	0.946	2.363
(1,2)	I	0.957	11.242	0.975	21.977	0.961	13.098	0.921	8.943	0.942	16.556	0.925	10.592	0.912	7.177
	II	0.961	6.637	0.976	10.351	0.966	7.517	0.938	7.429	0.954	15.540	0.944	9.356	0.924	6.099
	III	0.965	4.859	0.975	5.258	0.967	4.971	0.934	4.306	0.943	4.618	0.936	4.407	0.930	4.198
	IV	0.946	2.896	0.955	2.999	0.949	2.940	0.928	2.702	0.934	2.787	0.936	2.747	0.934	2.724
	V	0.952	1.673	0.953	1.710	0.956	1.699	0.945	1.624	0.950	1.660	0.950	1.651	0.948	1.650
(1,10)	I	0.963	5.098	0.969	7.328	0.967	5.373	0.935	4.086	0.947	6.156	0.933	4.451	0.919	3.588
	II	0.969	2.857	0.976	3.868	0.969	3.095	0.955	3.201	0.963	5.952	0.952	3.868	0.939	2.746
	III	0.958	2.398	0.960	2.482	0.956	2.411	0.938	2.170	0.940	2.229	0.940	2.183	0.933	2.138
	IV	0.952	1.455	0.954	1.470	0.951	1.458	0.940	1.393	0.945	1.405	0.948	1.397	0.946	1.393
	V	0.947	0.866	0.949	0.870	0.948	0.869	0.943	0.854	0.946	0.858	0.943	0.857	0.943	0.857
(2,10)	I	0.960	4.540	0.965	5.190	0.956	4.367	0.926	3.701	0.931	4.105	0.923	3.602	0.917	3.287
	II	0.954	2.639	0.957	2.890	0.953	2.635	0.935	2.822	0.937	3.278	0.934	2.815	0.925	2.471
	III	0.961	2.215	0.963	2.246	0.959	2.199	0.945	2.025	0.948	2.043	0.946	2.012	0.943	1.986
	IV	0.952	1.396	0.955	1.403	0.954	1.394	0.946	1.339	0.944	1.343	0.946	1.338	0.946	1.334
	V	0.946	0.829	0.949	0.831	0.949	0.829	0.945	0.815	0.945	0.817	0.945	0.817	0.946	0.816
(5,10)	I	0.951	3.645	0.960	3.794	0.953	3.439	0.911	3.054	0.919	3.123	0.909	2.923	0.910	2.798
	II	0.962	2.383	0.966	2.432	0.961	2.319	0.942	2.473	0.944	2.536	0.934	2.385	0.936	2.267
	III	0.963	1.962	0.963	1.971	0.960	1.944	0.943	1.801	0.943	1.804	0.942	1.787	0.942	1.772
	IV	0.951	1.245	0.952	1.247	0.953	1.240	0.944	1.190	0.943	1.190	0.944	1.187	0.943	1.185
	V	0.942	0.737	0.942	0.738	0.945	0.736	0.942	0.723	0.941	0.724	0.941	0.723	0.941	0.723

* I:(5,5), II:(5,10), III:(10,10), IV:(20,20), V: (50,50)

Table 2

Coverage probabilities (CP) and length of confidence interval (CI) of proposed 95% confidence intervals for the common gamma mean (2000 simulations) with two independent samples (K = 5).

(α1, α2)	Sizes*	GVK		GVC		GVW		MOVERK		MOVERC		MOVERW		MOVERboot
		CP	CI	CP	CI	CP	CI	CP	CI	CP	CI	CP	CI	CP	CI
μ = 1
(0.5,0.5,0.75,0.75,1)	VI	0.962	2.618	0.995	89.847	0.979	8.961	0.879	1.590	0.959	19.057	0.926	5.422	0.830	1.471
	VII	0.955	0.978	0.980	1.316	0.966	1.165	0.857	0.741	0.922	0.959	0.901	0.877	0.869	0.785
	VIII	0.931	0.561	0.961	0.639	0.956	0.620	0.874	0.475	0.913	0.533	0.908	0.520	0.898	0.510
	IX	0.959	0.687	0.984	2.241	0.975	1.184	0.927	1.196	0.966	15.958	0.949	4.277	0.913	1.213
	X	0.923	0.309	0.955	0.338	0.951	0.335	0.901	0.287	0.930	0.311	0.927	0.309	0.927	0.308
(0.5,1,2,5,10)	VI	0.970	0.903	0.987	3.417	0.976	1.300	0.901	0.609	0.944	2.858	0.922	1.146	0.863	0.551
	VII	0.961	0.396	0.969	0.425	0.964	0.407	0.906	0.322	0.924	0.346	0.916	0.335	0.899	0.322
	VIII	0.950	0.235	0.952	0.240	0.949	0.236	0.922	0.213	0.925	0.217	0.926	0.215	0.919	0.214
	IX	0.966	0.210	0.978	0.427	0.970	0.276	0.946	0.420	0.969	4.520	0.958	1.434	0.919	0.412
	X	0.956	0.135	0.960	0.137	0.957	0.136	0.947	0.130	0.950	0.132	0.948	0.132	0.947	0.131
(0.5,2,2,5,5)	VI	0.966	0.977	0.983	3.781	0.972	1.435	0.895	0.670	0.936	5.867	0.911	1.594	0.864	0.609
	VII	0.962	0.443	0.973	0.474	0.965	0.453	0.905	0.363	0.923	0.389	0.916	0.375	0.900	0.361
	VIII	0.945	0.268	0.951	0.274	0.950	0.270	0.916	0.241	0.923	0.246	0.922	0.244	0.917	0.243
	IX	0.957	0.276	0.976	0.602	0.965	0.368	0.944	0.504	0.966	5.257	0.954	1.571	0.922	0.495
	X	0.955	0.153	0.959	0.155	0.958	0.154	0.943	0.147	0.943	0.149	0.942	0.148	0.944	0.148
(1,2,2,5,5)	VI	0.971	0.916	0.983	1.551	0.972	0.954	0.904	0.629	0.924	0.887	0.901	0.654	0.864	0.537
	VII	0.960	0.426	0.966	0.442	0.958	0.425	0.915	0.354	0.918	0.363	0.911	0.354	0.901	0.346
	VIII	0.955	0.260	0.962	0.264	0.958	0.260	0.932	0.236	0.933	0.239	0.929	0.237	0.929	0.236
	IX	0.964	0.252	0.972	0.299	0.965	0.263	0.945	0.444	0.954	0.796	0.949	0.522	0.920	0.377
	X	0.945	0.151	0.948	0.152	0.945	0.151	0.939	0.145	0.941	0.146	0.938	0.146	0.940	0.146
(2,2,5,5,10)	VI	0.968	0.654	0.979	0.820	0.970	0.633	0.898	0.465	0.908	0.513	0.892	0.452	0.864	0.413
	VII	0.961	0.325	0.961	0.330	0.959	0.322	0.914	0.276	0.916	0.278	0.910	0.274	0.902	0.271
	VIII	0.966	0.202	0.967	0.204	0.964	0.202	0.941	0.186	0.940	0.187	0.941	0.186	0.936	0.186
	IX	0.961	0.180	0.967	0.188	0.959	0.180	0.936	0.302	0.948	0.345	0.932	0.300	0.915	0.267
	X	0.964	0.118	0.958	0.119	0.961	0.118	0.954	0.114	0.953	0.115	0.953	0.114	0.953	0.114
μ = 5
(0.5,0.5,0.75,0.75,1)	VI	0.968	12.835	0.994	396.747	0.982	43.412	0.882	7.805	0.964	126.406	0.924	32.435	0.835	7.311
	VII	0.953	4.907	0.979	6.617	0.973	5.857	0.860	3.728	0.924	4.794	0.903	4.393	0.867	3.940
	VIII	0.947	2.813	0.975	3.203	0.970	3.107	0.881	2.384	0.920	2.676	0.910	2.615	0.905	2.563
	IX	0.957	3.430	0.985	10.460	0.976	5.677	0.918	5.834	0.962	63.516	0.947	20.451	0.906	5.913
	X	0.928	1.552	0.961	1.695	0.960	1.679	0.899	1.437	0.930	1.557	0.931	1.546	0.928	1.542
(0.5,1,2,5,10)	VI	0.972	4.566	0.987	19.559	0.979	7.006	0.897	3.095	0.946	18.131	0.917	6.738	0.858	2.842
	VII	0.965	1.971	0.973	2.117	0.967	2.025	0.915	1.611	0.934	1.730	0.926	1.670	0.907	1.609
	VIII	0.949	1.177	0.954	1.202	0.953	1.188	0.923	1.067	0.927	1.089	0.926	1.080	0.928	1.074
	IX	0.957	1.047	0.976	2.291	0.968	1.384	0.939	2.091	0.959	24.026	0.949	7.508	0.919	2.071
	X	0.950	0.675	0.953	0.681	0.950	0.679	0.939	0.650	0.945	0.657	0.944	0.656	0.944	0.656
(0.5,2,2,5,5)	VI	0.966	4.907	0.984	16.560	0.974	6.850	0.894	3.356	0.936	19.666	0.912	6.848	0.860	3.047
	VII	0.960	2.229	0.973	2.381	0.968	2.276	0.897	1.822	0.918	1.953	0.906	1.882	0.892	1.812
	VIII	0.948	1.337	0.952	1.364	0.950	1.345	0.911	1.200	0.919	1.225	0.912	1.213	0.911	1.208
	IX	0.959	1.391	0.973	3.026	0.968	1.882	0.934	2.512	0.959	25.382	0.947	8.083	0.913	2.437
	X	0.952	0.766	0.952	0.775	0.949	0.772	0.939	0.735	0.943	0.744	0.942	0.742	0.943	0.741
(1,2,2,5,5)	VI	0.966	4.606	0.979	7.967	0.966	4.825	0.901	3.162	0.923	4.510	0.897	3.319	0.858	2.697
	VII	0.968	2.151	0.971	2.234	0.965	2.148	0.916	1.791	0.920	1.837	0.914	1.788	0.901	1.746
	VIII	0.969	1.304	0.972	1.322	0.967	1.305	0.941	1.186	0.941	1.197	0.940	1.187	0.935	1.183
	IX	0.956	1.267	0.972	1.499	0.961	1.321	0.933	2.208	0.948	4.063	0.936	2.631	0.913	1.882
	X	0.958	0.751	0.955	0.757	0.952	0.753	0.946	0.722	0.947	0.727	0.945	0.725	0.946	0.725
(2,2,5,5,10)	VI	0.965	3.276	0.970	4.063	0.960	3.143	0.901	2.312	0.912	2.527	0.891	2.243	0.870	2.063
	VII	0.964	1.626	0.964	1.648	0.960	1.610	0.923	1.379	0.922	1.389	0.919	1.368	0.916	1.351
	VIII	0.949	1.007	0.952	1.014	0.947	1.004	0.928	0.927	0.927	0.930	0.927	0.925	0.925	0.923
	IX	0.958	0.896	0.964	0.937	0.956	0.899	0.939	1.514	0.947	1.740	0.938	1.506	0.922	1.337
	X	0.951	0.591	0.953	0.593	0.950	0.591	0.941	0.572	0.942	0.573	0.942	0.572	0.941	0.573

* Sizes are VI:(5,5,5,5,5), VII:(10,10,10,10,10), VIII:(20,20,20,20,20), IX: (5,10,10,20,50), X: (50,50,50,50,50)

Table 1 presents simulated coverage probabilities (CP) and confidence interval lengths (CI) for K = 2. Overall speaking, three methods based on the generalized pivots (i.e. GV, GV, GV) maintains satisfactory coverage probabilities for all settings except that they might be slightly conservative at small sizes and GV was slightly liberal when (α1, α2) = (1, 2) at sample sizes (50, 50) and (α1, α2) = (0.5, 1) at sample sizes (20, 20). Among MOVER-type methods (i.e. MOVER, MOVER, MOVER, MOVER), MOVER performs the best while all of them are generally liberal when sample sizes are from (5, 5) to (20, 20). When sample sizes reach 50, all the proposed methods perform satisfactorily. The GV method provides shortest confidence intervals among three generalized pivots based methods, followed by GV. As sample sizes reach 20, all three methods (i.e. GV, GV, GV) are generally comparable. MOVER and MOVER provides shortest confidence intervals among MOVER-type methods. As sample sizes reach 20, all MOVER-type methods (i.e. MOVER, MOVER, MOVER, MOVER) are comparable in terms of length. When sample sizes reach 50, all the proposed methods generate confidence intervals with comparable length.

Table 2 presents simulated coverage probabilities (CP) and confidence interval lengths (CI) for K = 5. The three generalized pivots based methods methods generally maintains satisfactory coverage probabilities for all settings except that they tend to be slightly conservative at small sizes and GV is liberal at (α1, …, α5) = (0.5, 0.5, 0.75, 0.75, 1) with sizes (50, 50, 50, 50, 50). Among MOVER-type methods (i.e. MOVER, MOVER, MOVER, MOVER), MOVER performs the best while they are generally liberal when sample sizes are from (5, 5, 5, 5, 5) to (20, 20, 20, 20). When sample sizes reach 50, all methods perform satisfactorily. The GV method provides shortest confidence intervals among three generalized pivots based methods, followed by GV. As sample sizes reach 20, all three methods (i.e. GV, GV, GV) are comparable. MOVER provides shortest confidence intervals among three MOVER-type methods, followed by MOVER oot. As sample sizes reach 20, all four methods (i.e. MOVER, MOVER, MOVER, MOVER) are comparable. When sample sizes reach 50, all methods are generally comparable in terms of length.

In summary, generally we recommend GV, GV, GV methods over MOVER-type methods due to the fact that they can generate confidence intervals with satisfactory coverage probabilities even at smaller sizes. The MOVER-type methods are not recommended unless sample sizes are greater than or equal to 50. The large sample approach in (4) can severely underestimate the coverage probabilities, hence its results are not presented.

Data examples

In this section, we illustrate the proposed methods using two examples. Both datasets was analyzed in [11] for testing equality of gamma means, and it was concluded that the null hypothesis (equality of gamma means) can not be rejected. Therefore, in this paper, we use these two datasets to illustrate our proposed methods for estimating confidence intervals of the common gamma mean.

Example 1. Wright [29] reported ground water yield from two types of wells in southwestern Virginia. Table 3 presents this dataset which includes ground water yield data from 12 wells from valley underlain by unfractured rocks, and 13 wells by fractured rocks. It has been argued that gamma distribution is appropriate to fit the data in each sample, and the test for equality of means [11] concluded that the means of water yields from two types of wells are the same. The estimated parameters are: α1 = 0.4342, , α2 = 1.1854, . The estimated 95% confidence intervals for the common gamma mean by all the proposed methods are presented in Table 4. Our simulation study demonstrate that MOVER-type methods could be liberal at sample sizes (10, 10). Give the sample sizes as 12 and 13 in this application, the confidence intervals by GV, GV, GV methods are recommended, and among them the GV method has the shortest length.

Table 3

Virginia ground water well yields data (in gal/min/ft) [29].

Without fractures	with fractures
0.001, 0.003, 0.007, 0.020	0.020, 0.031, 0.085, 0.013
0.030, 0.040, 0.041, 0.077	0.160, 0.160, 0.180, 0.300
0.100, 0.454, 0.490, 1.020	0.400, 0.440, 0.510, 0.720, 0.950

Table 4

Estimated confidence interval for Virginia ground water well yields data (in gal/min/ft).

method	lower	upper	LCI
GV K	0.140	0.491	0.351
GV C	0.159	0.576	0.417
GV W	0.153	0.574	0.421
MOVER K	0.169	0.459	0.289
MOVER C	0.178	0.548	0.370
MOVER W	0.175	0.530	0.355
MOVER boot	0.175	0.490	0.314

Example 2. Table 5 presents a dataset of chloride concentration in spring water samples from two types of rocks in Sierra Nevada, California and Nevada [30]. It has been argued that gamma distribution is appropriate to fit the data in each sample, and testing equality of means [11] concluded that the means of chloride concentration from two types of rocks are the same. The estimated parameters are: α1 = 0.7594, , α2 = 1.1359, . The estimated 95% confidence intervals for the common gamma mean by all the proposed methods are presented in Table 6. Given sample sizes 18 and 17 and parameter estimates, the confidence interval estimated by GV is most recommenced in practice.

Table 5

Chloride concentration (in mg/litre) in water data. [30].

Granodiorite	Quartz Monzonite
6.0, 0.5, 0.4, 0.7, 0.8, 6.0, 5.0, 0.6, 1.2	1.0, 0.2, 1.2, 1.0, 0.3, 0.1, 0.1, 0.4, 3.2
1.0, 0.2, 1.2, 1.0, 0.3, 0.1, 0.1, 0.4, 3.2	0.3, 0.4, 1.8, 0.9, 0.1, 0.2, 0.3, 0.5

Table 6

Estimated confidence intervals and lengths for the common mean for Chloride concentration (in mg/litre) in water.

method	lower	upper	length
GV K	0.524	1.366	0.842
GV C	0.555	1.482	0.928
GV W	0.547	1.455	0.908
MOVER K	0.543	1.251	0.707
MOVER C	0.569	1.317	0.749
MOVER W	0.571	1.317	0.746
MOVER boot	0.567	1.310	0.743

Summary and discussion

Gamma distribution plays an important role in practice. When the result of testing equality of several gamma means is not significant, it is customary that we need to make inference about the common gamma mean. While the standard large sample methods exist, small sample inference for the common gamma mean has not been explored. In this article, we focus on accurate confidence interval estimation for the common gamma mean based on several independent gamma samples using the concepts of generalized pivots and the method of MOVER. Via a comprehensive simulation study, we discovered that the proposed methods based on generalized pivots can generally provide satisfactory confidence intervals with consistent performance despite parameter settings and sample sizes. The MOVER-type methods can be liberal for certain scenarios, especially when sample sizes are small.

The proposed methods are easy to implement. The R program is available at li.yan@roswellpark.org.

Due to the popularity of gamma distribution in applied fields, we expect the proposed methods have wide applicability in practice where right-skewed data are often observed.

11 Jan 2022

A rebuttal letter that responds to each point raised by the academic editor and reviewer(s). You should upload this letter as a separate file labeled 'Response to Reviewers'.

A marked-up copy of your manuscript that highlights changes made to the original version. You should upload this as a separate file labeled 'Revised Manuscript with Track Changes'.

An unmarked version of your revised paper without tracked changes. You should upload this as a separate file labeled 'Manuscript'.

If you would like to make changes to your financial disclosure, please include your updated statement in your cover letter. Guidelines for resubmitting your figure files are available below the reviewer comments at the end of this letter.

If applicable, we recommend that you deposit your laboratory protocols in protocols.io to enhance the reproducibility of your results. Protocols.io assigns your protocol its own identifier (DOI) so that it can be cited independently in the future. For instructions see: https://journals.plos.org/plosone/s/submission-guidelines#loc-laboratory-protocols. Additionally, PLOS ONE offers an option for publishing peer-reviewed Lab Protocol articles, which describe protocols hosted on protocols.io. Read more information on sharing protocols at https://plos.org/protocols?utm_medium=editorial-email&utm_source=authorletters&utm_campaign=protocols.

We look forward to receiving your revised manuscript.

Kind regards,

Miguel A. Fernández, Ph.D.

Academic Editor

PLOS ONE

Journal Requirements:

When submitting your revision, we need you to address these additional requirements.

1. Please ensure that your manuscript meets PLOS ONE's style requirements, including those for file naming. The PLOS ONE style templates can be found at

https://journals.plos.org/plosone/s/file?id=wjVg/PLOSOne_formatting_sample_main_body.pdf and

https://journals.plos.org/plosone/s/file?id=ba62/PLOSOne_formatting_sample_title_authors_affiliations.pdf

2. We suggest you thoroughly copyedit your manuscript for language usage, spelling, and grammar. If you do not know anyone who can help you do this, you may wish to consider employing a professional scientific editing service.

Whilst you may use any professional scientific editing service of your choice, PLOS has partnered with both American Journal Experts (AJE) and Editage to provide discounted services to PLOS authors. Both organizations have experience helping authors meet PLOS guidelines and can provide language editing, translation, manuscript formatting, and figure formatting to ensure your manuscript meets our submission guidelines. To take advantage of our partnership with AJE, visit the AJE website (http://aje.com/go/plos) for a 15% discount off AJE services. To take advantage of our partnership with Editage, visit the Editage website (www.editage.com) and enter referral code PLOSEDIT for a 15% discount off Editage services.  If the PLOS editorial team finds any language issues in text that either AJE or Editage has edited, the service provider will re-edit the text for free.

Upon resubmission, please provide the following:

The name of the colleague or the details of the professional service that edited your manuscript

A copy of your manuscript showing your changes by either highlighting them or using track changes (uploaded as a *supporting information* file)

A clean copy of the edited manuscript (uploaded as the new *manuscript* file)

[Note: HTML markup is below. Please do not edit.]

Reviewers' comments:

Reviewer's Responses to Questions

Comments to the Author

1. Is the manuscript technically sound, and do the data support the conclusions?

The manuscript must describe a technically sound piece of scientific research with data that supports the conclusions. Experiments must have been conducted rigorously, with appropriate controls, replication, and sample sizes. The conclusions must be drawn appropriately based on the data presented.

Reviewer #1: Yes

Reviewer #2: Partly

**********

2. Has the statistical analysis been performed appropriately and rigorously?

Reviewer #1: Yes

Reviewer #2: Yes

**********

3. Have the authors made all data underlying the findings in their manuscript fully available?

The PLOS Data policy requires authors to make all data underlying the findings described in their manuscript fully available without restriction, with rare exception (please refer to the Data Availability Statement in the manuscript PDF file). The data should be provided as part of the manuscript or its supporting information, or deposited to a public repository. For example, in addition to summary statistics, the data points behind means, medians and variance measures should be available. If there are restrictions on publicly sharing data—e.g. participant privacy or use of data from a third party—those must be specified.

Reviewer #1: Yes

Reviewer #2: Yes

**********

4. Is the manuscript presented in an intelligible fashion and written in standard English?

PLOS ONE does not copyedit accepted manuscripts, so the language in submitted articles must be clear, correct, and unambiguous. Any typographical or grammatical errors should be corrected at revision, so please note any specific errors here.

Reviewer #1: Yes

Reviewer #2: Yes

**********

5. Review Comments to the Author

Please use the space provided to explain your answers to the questions above. You may also include additional comments for the author, including concerns about dual publication, research ethics, or publication ethics. (Please upload your review as an attachment if it exceeds 20,000 characters)

Reviewer #1: The author has addressed the problem of estimating the common mean of several gamma distributions. He has proposed several confidence intervals (CIs) based on the generalized variable approach, MOVER and likelihood approach. The proposed CIs are evaluated and compared with respect to coverage probabilities and expected widths. In general, the paper is easy to read and I have the following comments and suggestions.

Reviewer #2: The paper considers constructing confidence interval for the common mean of gamma distributed samples and the author proposes two category of methods which are validated by using simulations. Although the numerical results are very good, I have a concern of the theoretical foundation of the proposed method. Below are my major comments.

1. What is the rationale behind the constructed GPQs (11) and (12)? Are they valid GPQs? Are there any theoretical guarantees of using them in constructing the confidence intervals?

2. Line 113, Step 3. Is it Eq (8) or Eq (3)?

3. The common mean problem of the gamma distribution has been well identified in the literature, as the author claimed. There must be some existing methods in constructing the confidence interval. The author needs to compare the proposed methods with the existing ones using simulations.

4. For the examples, why is it important to assume a common mean for the two samples? It seems that confidence intervals can be well constructed for each individual sample.

**********

6. PLOS authors have the option to publish the peer review history of their article (what does this mean?). If published, this will include your full peer review and any attached files.

If you choose “no”, your identity will remain anonymous but your review may still be made public.

Do you want your identity to be public for this peer review? For information about this choice, including consent withdrawal, please see our Privacy Policy.

Reviewer #1: Yes: Kalimuthu Krishnamoorthy

Reviewer #2: No

[NOTE: If reviewer comments were submitted as an attachment file, they will be attached to this email and accessible via the submission site. Please log into your account, locate the manuscript record, and check for the action link "View Attachments". If this link does not appear, there are no attachment files.]

While revising your submission, please upload your figure files to the Preflight Analysis and Conversion Engine (PACE) digital diagnostic tool, https://pacev2.apexcovantage.com/. PACE helps ensure that figures meet PLOS requirements. To use PACE, you must first register as a user. Registration is free. Then, login and navigate to the UPLOAD tab, where you will find detailed instructions on how to use the tool. If you encounter any issues or have any questions when using PACE, please email PLOS at figures@plos.org. Please note that Supporting Information files do not need this step.

Submitted filename: Report.pdf

Click here for additional data file.

3 Feb 2022

The response to reviewers were uploaded in pdf files (Responses_to_reviewer_1.pdf and Responses_to_reviewer_2.pdf) for proper format

Submitted filename: Responses_to_reviewer_2.pdf

Click here for additional data file.

10 Mar 2022

A rebuttal letter that responds to each point raised by the academic editor and reviewer(s). You should upload this letter as a separate file labeled 'Response to Reviewers'.

A marked-up copy of your manuscript that highlights changes made to the original version. You should upload this as a separate file labeled 'Revised Manuscript with Track Changes'.

An unmarked version of your revised paper without tracked changes. You should upload this as a separate file labeled 'Manuscript'.

If you would like to make changes to your financial disclosure, please include your updated statement in your cover letter. Guidelines for resubmitting your figure files are available below the reviewer comments at the end of this letter.

If applicable, we recommend that you deposit your laboratory protocols in protocols.io to enhance the reproducibility of your results. Protocols.io assigns your protocol its own identifier (DOI) so that it can be cited independently in the future. For instructions see: https://journals.plos.org/plosone/s/submission-guidelines#loc-laboratory-protocols. Additionally, PLOS ONE offers an option for publishing peer-reviewed Lab Protocol articles, which describe protocols hosted on protocols.io. Read more information on sharing protocols at https://plos.org/protocols?utm_medium=editorial-email&utm_source=authorletters&utm_campaign=protocols.

We look forward to receiving your revised manuscript.

Kind regards,

Miguel A. Fernández, Ph.D.

Academic Editor

PLOS ONE

[Note: HTML markup is below. Please do not edit.]

Reviewers' comments:

Reviewer's Responses to Questions

Comments to the Author

1. If the authors have adequately addressed your comments raised in a previous round of review and you feel that this manuscript is now acceptable for publication, you may indicate that here to bypass the “Comments to the Author” section, enter your conflict of interest statement in the “Confidential to Editor” section, and submit your "Accept" recommendation.

Reviewer #1: All comments have been addressed

Reviewer #2: (No Response)

**********

2. Is the manuscript technically sound, and do the data support the conclusions?

The manuscript must describe a technically sound piece of scientific research with data that supports the conclusions. Experiments must have been conducted rigorously, with appropriate controls, replication, and sample sizes. The conclusions must be drawn appropriately based on the data presented.

Reviewer #1: Yes

Reviewer #2: Partly

**********

3. Has the statistical analysis been performed appropriately and rigorously?

Reviewer #1: Yes

Reviewer #2: Yes

**********

4. Have the authors made all data underlying the findings in their manuscript fully available?

The PLOS Data policy requires authors to make all data underlying the findings described in their manuscript fully available without restriction, with rare exception (please refer to the Data Availability Statement in the manuscript PDF file). The data should be provided as part of the manuscript or its supporting information, or deposited to a public repository. For example, in addition to summary statistics, the data points behind means, medians and variance measures should be available. If there are restrictions on publicly sharing data—e.g. participant privacy or use of data from a third party—those must be specified.

Reviewer #1: Yes

Reviewer #2: Yes

**********

5. Is the manuscript presented in an intelligible fashion and written in standard English?

PLOS ONE does not copyedit accepted manuscripts, so the language in submitted articles must be clear, correct, and unambiguous. Any typographical or grammatical errors should be corrected at revision, so please note any specific errors here.

Reviewer #1: Yes

Reviewer #2: Yes

**********

6. Review Comments to the Author

Please use the space provided to explain your answers to the questions above. You may also include additional comments for the author, including concerns about dual publication, research ethics, or publication ethics. (Please upload your review as an attachment if it exceeds 20,000 characters)

Reviewer #1: The tables need ti be reformatted. Columns of coverage probabilities and corresponding expected widths are not in alignment. Sample sizes are not reported for the case of four populations.

Reviewer #2: I appreciate the author's efforts in addressing my comments but I still have concerns regarding some responses.

Firstly, in my previous Comment#1, what I asked is whether the constructed GPQs are valid. There are two conditions for a valid GPQ and the author needs to check them. In addition, it is good to see the superb performance of the GPQ method, but is there any theoretical guarantee? If no, probably the good performance is just because of the simulation settings. The author needs to provide some insights in using GPQs.

Secondly, in the response to my Comment#3, the authors mentioned all other methods do not perform well at finite sample sizes. I do not see the rational between the two sentences ""In page 2, some large sample approach was presented" and "However, as expected, its confidence intervals can be severely liberal". The authors clearly do not conduct numerical simulations so what is the evidence of "expectation"? I would suggest adding simulation results from other methods to the paper to better highlight the contribution.

**********

7. PLOS authors have the option to publish the peer review history of their article (what does this mean?). If published, this will include your full peer review and any attached files.

If you choose “no”, your identity will remain anonymous but your review may still be made public.

Do you want your identity to be public for this peer review? For information about this choice, including consent withdrawal, please see our Privacy Policy.

Reviewer #1: No

Reviewer #2: No

[NOTE: If reviewer comments were submitted as an attachment file, they will be attached to this email and accessible via the submission site. Please log into your account, locate the manuscript record, and check for the action link "View Attachments". If this link does not appear, there are no attachment files.]

While revising your submission, please upload your figure files to the Preflight Analysis and Conversion Engine (PACE) digital diagnostic tool, https://pacev2.apexcovantage.com/. PACE helps ensure that figures meet PLOS requirements. To use PACE, you must first register as a user. Registration is free. Then, login and navigate to the UPLOAD tab, where you will find detailed instructions on how to use the tool. If you encounter any issues or have any questions when using PACE, please email PLOS at figures@plos.org. Please note that Supporting Information files do not need this step.

22 Apr 2022

The response to reviewers were uploaded in pdf files (Responses to reviewer 1_R2.pdf

and Responses to reviewer 2_R2.pdf) for proper format

Submitted filename: Responses to reviewer 2_R2.pdf

Click here for additional data file.

2 Jun 2022

Confidence interval estimation of the common mean of several gamma populations

PONE-D-21-27089R2

Dear Dr. Yan,

We’re pleased to inform you that your manuscript has been judged scientifically suitable for publication and will be formally accepted for publication once it meets all outstanding technical requirements.

Within one week, you’ll receive an e-mail detailing the required amendments. When these have been addressed, you’ll receive a formal acceptance letter and your manuscript will be scheduled for publication.

An invoice for payment will follow shortly after the formal acceptance. To ensure an efficient process, please log into Editorial Manager at http://www.editorialmanager.com/pone/, click the 'Update My Information' link at the top of the page, and double check that your user information is up-to-date. If you have any billing related questions, please contact our Author Billing department directly at authorbilling@plos.org.

If your institution or institutions have a press office, please notify them about your upcoming paper to help maximize its impact. If they’ll be preparing press materials, please inform our press team as soon as possible -- no later than 48 hours after receiving the formal acceptance. Your manuscript will remain under strict press embargo until 2 pm Eastern Time on the date of publication. For more information, please contact onepress@plos.org.

Kind regards,

Miguel A. Fernández, Ph.D.

Academic Editor

PLOS ONE

Additional Editor Comments (optional):

Reviewers' comments:

Reviewer's Responses to Questions

Comments to the Author

1. If the authors have adequately addressed your comments raised in a previous round of review and you feel that this manuscript is now acceptable for publication, you may indicate that here to bypass the “Comments to the Author” section, enter your conflict of interest statement in the “Confidential to Editor” section, and submit your "Accept" recommendation.

Reviewer #1: All comments have been addressed

Reviewer #2: All comments have been addressed

**********

2. Is the manuscript technically sound, and do the data support the conclusions?

The manuscript must describe a technically sound piece of scientific research with data that supports the conclusions. Experiments must have been conducted rigorously, with appropriate controls, replication, and sample sizes. The conclusions must be drawn appropriately based on the data presented.

Reviewer #1: Yes

Reviewer #2: Yes

**********

3. Has the statistical analysis been performed appropriately and rigorously?

Reviewer #1: Yes

Reviewer #2: Yes

**********

4. Have the authors made all data underlying the findings in their manuscript fully available?

The PLOS Data policy requires authors to make all data underlying the findings described in their manuscript fully available without restriction, with rare exception (please refer to the Data Availability Statement in the manuscript PDF file). The data should be provided as part of the manuscript or its supporting information, or deposited to a public repository. For example, in addition to summary statistics, the data points behind means, medians and variance measures should be available. If there are restrictions on publicly sharing data—e.g. participant privacy or use of data from a third party—those must be specified.

Reviewer #1: Yes

Reviewer #2: Yes

**********

5. Is the manuscript presented in an intelligible fashion and written in standard English?

PLOS ONE does not copyedit accepted manuscripts, so the language in submitted articles must be clear, correct, and unambiguous. Any typographical or grammatical errors should be corrected at revision, so please note any specific errors here.

Reviewer #1: Yes

Reviewer #2: Yes

**********

6. Review Comments to the Author

Please use the space provided to explain your answers to the questions above. You may also include additional comments for the author, including concerns about dual publication, research ethics, or publication ethics. (Please upload your review as an attachment if it exceeds 20,000 characters)

Reviewer #1: As noted in the paper, inferences for a single gamma

mean or for comparison of several gamma means have been well studied. However, research on the common mean of several gamma

populations are sparse. This paper addresses the problem of confidence interval

estimation of the common mean of several gamma populations using the fiducial inference. Simulation studies have been carried out to judge the accuracy of the proposed methods

I have no further comments. The manuscript maybe accepted as it is.

Reviewer #2: (No Response)

**********

7. PLOS authors have the option to publish the peer review history of their article (what does this mean?). If published, this will include your full peer review and any attached files.

If you choose “no”, your identity will remain anonymous but your review may still be made public.

Do you want your identity to be public for this peer review? For information about this choice, including consent withdrawal, please see our Privacy Policy.

Reviewer #1: Yes: Kalimuthu Krishnamoorthy

Reviewer #2: No

**********

10 Jun 2022

PONE-D-21-27089R2

Confidence Interval Estimation of the Common Mean of Several Gamma Populations

Dear Dr. Yan:

I'm pleased to inform you that your manuscript has been deemed suitable for publication in PLOS ONE. Congratulations! Your manuscript is now with our production department.

If your institution or institutions have a press office, please let them know about your upcoming paper now to help maximize its impact. If they'll be preparing press materials, please inform our press team within the next 48 hours. Your manuscript will remain under strict press embargo until 2 pm Eastern Time on the date of publication. For more information please contact onepress@plos.org.

If we can help with anything else, please email us at plosone@plos.org.

Thank you for submitting your work to PLOS ONE and supporting open access.

Kind regards,

PLOS ONE Editorial Office Staff

on behalf of

Dr Miguel A. Fernández

Academic Editor

PLOS ONE