The rural areas (14 rural thanas) of Chittagong district was considered as the study area. Chittagong district is one of 64 districts in Bangladesh. It has, however, almost similar type of homestead condition, household occupation, home garden with tree and bamboo plantation, etc. as in other rural areas in the country.
A target population was formed in the study area. A preliminary survey was conducted with a sample of 45 households. A field sheet containing homestead area, compact tree garden area, diameter at breast height, height measurement of trees in both compact and scattered areas, number of bamboos with culms, general information, economic status, etc., of a household was developed. A total of 11 major tree and 3 common bamboo species were selected for data collection. A number of many minor tree species was also grouped as others. Similarly a number of a few uncommon bamboo species was grouped as others.
A total of 7 parameters and 5 variables were considered for the study. The seven parameters of total homestead area (ha), total area (ha) of compact tree garden, total volume(m3 ) of trees, total volume (m3 ) of compact trees, total number of bamboos, volume (m3) of compact trees per ha and number of households with bamboo plantation were considered. The five variables of area of compact tree garden (x2), homestead area (x4), volume of standing trees (y2), volume of standing compact trees (y3) and number of standing bamboos (y4) of a household were also considered.
The target population of 900 households was estimated by a typical formula (Chacko, 1965) on the basis of variable y2 with 12% margin of error. Homestead area and area under compact tree garden of a household were measured by a long tape in ha and recorded in the field sheet. The diameter at breast height (dbh) measurement of all standing trees (dbh larger than 20 cm) except narikel (Cocos nucifera) and tal (Borassus fiabellfer) was taken by diameter tape. Volume of all trees was estimated by local volume tables (Aleem, 1981; Islam, 1984; Islam, 1988). A total of 22081 trees and a total of 53229 bamboos of selected species were tallied in 1996 and recorded in the field sheet. Although data of trees and bamboos were collected a long ago, it did not hamper to compare estimated values with enumerated values for selection of a best sampling method. The present paper deals with selection of a best sampling method rather than estimating parameters. So, it may not be necessary for recent data. Recent data serve the purpose of up to date estimation.
The five variables are generated through computer. The complete enumeration is carried out and the seven parameters are calculated. Cluster Sampling, Two Stage Sampling and Stratified Two Stage Sampling (Self-Weighted) were chosen for field testing in order to recommend appropriate sampling design. The chosen sampling designs were applied on the target population for estimating the parameters. Stratification was done on the basis of location, land use system and homestead wood and bamboo resources so that each stratum was internally homogeneous. The 14 rural thanas were thus stratified as stratum-I (Mirsarai + Sitakunda), stratum-II (Sandwip), stratum-III (Hathazari+ Fatikchari), stratum-IV (Rawzan+ Rangunia), stratum-V (Anwara+ Patiya+ Boakhali), stratum-VI (Satkania+Lohagara), stratum-VII (Banskhali+Chandanaish). The 22 villages and 900 households were distributed among the 7 strata. A sample of 250 households was considered to estimate the parameters by the chosen sampling designs. The population total and ratio were estimated by the formula of total estimate and ratio estimate (Obaidullah, 1980). The population proportion was estimated by the formula of proportion estimate (Islam, 1994). The study is to compare the estimates done by chosen sampling designs with the enumerated values of the parameters in respect of percentage standard error (%SE), percentage accuracy (%AC), percentage bias (%bias) and 95% confidence interval. The cost of the survey was not considered in the study. The percentage standard error of the estimate was estimated by (SE/estimate) 100. Similarly the percentage bias was estimated. The percentage accuracy was estimated by 100-{(population value – estimate)/population value} 100. The 95% confidence internal was estimated by usual formula. The sample of 14 villages (600 households) and 250 households selected randomly from seven strata were considered as first stage and second stage units respectively.