ffi ciency of Alodat Sample Selection Procedure over Sen-Midzuno and Yates-Grundy Draw by Draw under Unequal Probability Sampling without Replacement Sample Size 2

Received: November 8, 2010 Accepted: December 7, 2010 doi:10.5539/jmr.v3n2p113 Abstract This paper compare the efficiency of Alodat sample selection procedure over Sen Midzuno and Yates Grundy draw by draw using Yates Grundy estimator under unequal probability sampling without replacement sample size 2, carried out using the data from the 2008 Demographic and health survey in Nigeria. We studied the distribution of pregnant women age 15 49, and children under age five in Nigeria, who use mosquito nets as a means of preventing malaria. These data sets are: (1) the number of pregnant women age 15 49 who slept under mosquito nets the night before the survey, and (2) the number of children under age five who slept under mosquito nets the night before the survey. (1) and (2) above are the variables of interest. The data were collected based on the six geo-political zones in Nigeria [i.e. South South, South West, South East, North West, North East, North Central]. The auxiliary variable is the number of Local Government in each geo-political zone in Nigeria. The Yates Grundy estimate obtained using Alodat sample selection is more efficient than using Sen Midzuno and Yates Grundy selection procedures.


Introduction
For the purpose of this research work, interest is on the unequal probability sampling (UPS), which is also called probability proportional to size (PPS) sampling.It can be done with and without replacement of the unit selected.The abundance of works about sampling designs without replacement with fixed sample size is quite disconcerting.No less than fifty sampling procedures are presented in the famous paper of HANIF and BREWER [1980].Several other recent papers deal with this matter as well.Among them, one can cite the CHAO updating procedure [1982] and the DEVILLE method [1992].The study of UPS is a big task.Several people have studies the subject.They include; Horvitz and Thompson (1952), Yates and Grundy (1953), Sen (1953), Durbin (1953), Rao, Hartley and Cochran (1962), Hansen andHurwitz (1943), etc. Horvitz andThompson (1952) were the first to provide a complete theoretical frame work for UPS without replacement.There have been a lot of improvement on UPS over the years.The search for a sampling design with unequal probabilities is a relatively open problem.A 'good' solution should however respect the following properties.
1.The procedure should be exact in the sense that the units should be selected exactly with probabilities equaling (1) 2. The procedure should be general i.e. it should be possible to apply it to any set of first -order inclusion probabilities fixed a priori, which satisfy the relation (1).
3. The p(s) obtained by means of the sampling method should not depend on the order of the units on the data file.
4. The joint inclusion probabilities should be easy to compute without examining all the probabilities p(s).
5. The joint inclusion probabilities should be strictly positive.
6.The joint inclusion probabilities should verify the Yates-Grundy condition: Here, in dealing with UPS, we studied the distribution of pregnant women age 15 -49 and children under age five in all house holds in the six geo-political zones in Nigeria, who slept under a mosquito net (treated or untreated), under an ever -treated mosquito net, and under an insecticide -treated net.This is carried out using the 2008 Demographic and Health Survey in Nigeria, which was implemented by Nigerian National Population Commission (NPC).
Malaria is endemic throughout Nigeria.The National Malaria Control Strategic Plan (NMCSP) addresses National Health Development priorities including the Roll Back Malaria Goals, and the Millennium Development Goals.
The NMCSP includes the following priorities: to reduce malaria related mortality, to reduce malaria parasite prevalence in children under age five, to increase ownership and use of insecticide -treated net and long -lasting insecticidal nets, to introduce and scale up indoor residual spraying, to increase the use of diagnostic tests for fever patients, to improve appropriate and timely treatment of malaria, and to increase coverage of intermittent preventive treatment of malaria during pregnancy.
It is restricted to the use of mosquito nets as a way of preventing malaria among pregnant women age 15 -49 years and children under age five years in Nigeria.
The main objective of carrying out this research is to determine the best selection procedure out of the three procedures under consideration for estimating the population of pregnant women age 15 -49 and children under age five in Nigeria, who use mosquito nets to prevent malaria.

Data Used
The data used for this research is secondary data on the number of Local Government in each geo -political zone in Nigeria collected from the National Population Commission, Ibadan.Nigeria on the use of mosquito nets extracted from the 2008 Demographic and Health Survey in Nigeria as shown in tables 1 -3(b).

Sample Selection Procedures Considered
(i) Sen -Midzuno: This selection procedure is reported by Horvitz and Thompson (1952) and is applicable for a sample of any size.This selection procedure is stated as -Select first unit with P i probability -Select a sample of size n − 1 from remaining units with equal probability and without replacement.The quantities i and ij for this selection procedure are given as: therefore, for a sample of size n = 2 and N = 6, which is the basis of this research, (ii) Yates -Grundy Draw -by -Draw: This selection procedure was developed by Yates and Grundy (1953) and also reported by Durbin (1953) and Hajek (1964).The selection procedure is stated as: -Select first unit with probability proportional to size -Select second unit with probability proportional to size of remaining units.The quantities π i and π i j for this selection procedure are given as: where (iii) Alodat The procedure is given for the use with Horvitz and Thompson estimator with sample size 2.
-Select first unit with probability proportional to.Vol. 3, No. 2; May 2011 and without replacement.
-Select second unit with probability The first inclusion probability is. where Π i j = P i P( j/1) + P j P(i/ j) (16) Our intension is to apply this sample selection procedure to Yate and Grundy estimator instead of the Horvitz and Thompson estimator with sample size 2 which was initially proposed for.

Estimator Considered
Yate and Grundy Estimator.Yates and Grundy (1953) suggested an alternative estimator of variance which is believed to be "less often negative than Horvitz and Thompson estimator of variance" and its unbiased estimator of variance is.
In other to find the estimates of the three selection procedures under consideration and Yates and Grundy estimator, we then have these summary of inclusion probabilities shown in table 4 from the data set to be used in other to make our computations easier.The following conditions are met

Summary of Results
Using the inclusion probabilities shown in table 4, we then compute the variance of Yates and Grundy estimator based on the three selection procedures for the two data sets used as shown in table 5.

Discussion of Results
Considering the analysis and estimates obtained, Alodat selection procedure gave the least variance under both data sets.This is followed by Yates and Grundy Draw -by Draw procedure.Sen -Midzuno procedure gave the highest variance using both data sets.Alodat ( 2009) selection procedure is the best selection procedure in this paper.

Conclusion
Finally, for estimating the number of pregnant women age 15 -49 and children under five years in the six geo -political zones in Nigeria, who use mosquito nets as a means of preventing malaria, using unequal probability sampling without replacement on Yates and Grundy estimator, among all the three selection procedures considered in this paper, Alodat (2009) selection procedure is the most efficient of them all.Yates and Grundy 20, 385, 287.82 788, 188.9463 Yates and Grundy(1953) Yates and Grundy 13,601,913.64 629,997.7596

Table 1 .
Distribution of the number of Local Government in each Geo -Political Zones in Nigeria Zones Total no. of Local Govt. in each Zone (X i )

Table 3b .
Distribution of the total number of pregnant women age 15 − 49 who slept under any net, an ever-treated net or an insecticide treated net [Data 2].

Table 5 .
Variance of Yates and Grundy estimator based on the three selection procedures.