Sample Sizes and Error Margins. An indicator of data quality is the standard error of the estimate, on which the margin for sampling error is based. As survey statistics are mostly proportions, the key measure of data precision is the standard error of a proportion taken from a sample. It is computed as follows:
Where Z, at 95% confidence level is 1.96; p is the sample proportion estimate and n is the sample size. The overall sample size of 1,200 voting-age adults gives a maximum error margin of ± 2.83 % at the 95% confidence level, assuming a simple random sampling design. The sampling error is at its highest when the true proportion being estimated is close to 50%.
+_ Z
*
p(1-pnThe following approximate 95%-confidence margins for sampling error should be made when aggregating data at various levels:
Sample Size Error margin
Philippines 1200 ±3%
National Capital Region 300 ±6%
Balance Luzon 300 ±6%
Visayas 300 ±6%
Mindanao 300 ±6%
However, somewhat higher error margins should be expected since multi-stage cluster sampling was used; this design-effect is not readily measurable through established statistical software.
Sampling scheme. The Philippines was divided into four study areas: National Capital Region (NCR), Balance Luzon, Visayas, and Mindanao. The sample size for each of the four study areas is 300 voting-age adults.
Multi-stage probability sampling was used in the selection of sample spots. The allocation of sample units in each stage was as follows:
Sample Sample Probability
Prov. Mun. Spots Respondents National Capital Region -- 17 60 300 Balance Luzon 10 15 60 300
Visayas 5 15 60 300
Mindanao 6 15 60 300
--- --- --- ---
21 62 240 1200
For the National Capital Region
Stage 1. Selection of Sample Spots (Barangays)
For NCR's first stage, 60 barangays are distributed among the 17 NCR cities and municipalities in such a way that each city/municipality was assigned a number of barangays that was roughly proportional to its population size. An additional provision was that each municipality must receive at least one barangay.
Barangays were then selected from within each municipality with probability proportional to size (PPS).
Stage 2. Selection of Sample Households
In each sample barangay, five households were established by systematic sampling.
Designated starting points were randomly assigned - it was either: 1) a
municipal/barangay hall, 2) a school, 3) the barangay captain's house, or 4) a church/chapel/mosque. A random start from 1-6 was also randomly generated for each spot. Thus, if a particular spot has a random start of 4, the first sample household should be the 4th household from the designated starting point.
Subsequent sample households were chosen using a fixed interval of 5 households in between the sampled ones; i.e. every 6th household was sampled.
Stage 3. Selection of Sample Adult
For the third stage, in each selected household, a respondent is randomly chosen among the household members who were 18 years of age and older, using a probability selection table. In selecting the probability respondent of a household, only male family members were pre-listed in the probability selection table of odd-numbered questionnaires; only female family members were pre-listed for even-numbered questionnaires. A respondent not contacted during the first attempt was visited for a second time. If the respondent remained unavailable, or in cases where there was no qualified probability respondent of a given gender, the interval sampling of households would continue until five sample respondents were identified.
For the rest of the Philippines
Stage 1. Allocation and Selection of Sample Provinces
Balance Luzon was further divided into 5 regions: Region I, CAR + Region II, Region III, Region IV and Region V; Visayas into 3 regions: Region VI, Region VII and Region VIII; and Mindanao into 6 regions; Region IX, Region X, CARAGA, Region XI, Region XII and ARMM.
Using probability proportional to population size (PPS) of the region, the allocation of 10 provinces in Luzon, 5 in Visayas and 6 in Mindanao were as follows:
LUZON VISAYAS MINDANAO
Region I 1 Region VI 2 Region IX 1
CAR/REG II 1 Region VII 1 Reg X 1 Region III 2 Region VIII 1 CARAGA 1
Region IV 3 Non-quota 1 Region XI 1
Region V 1 Region XII 1
Non-quota 2 ARMM 1
---- ---- ----
TOTAL 10 5 6
The non-quota provinces were selected without replacement using probability proportional to their remainders. The remainders are fractions derived when the proportion of the regions (based on their respective study area) are multiplied by 10
for Luzon, and 5 for Visayas and 6 for Mindanao. For instance, if 1.45 is obtained for Region I, then 1 province is assigned to this region and remaining fraction of 0.45 is included for the allocation of the non-quota province.
Given the target number of provinces for each region, sample provinces were then selected by PPS, without replacement. An additional provision is that each region must receive at least one province.
Stage 2. Allocation and selection of sample municipalities
Within each study area, 15 municipalities were allocated among the sample provinces. 15 was multiplied by the proportion of the provinces. The resulting integers became the number of municipalities in that province. If there were remaining municipalities to be allocated, they were distributed using probability proportional to the remainders.
Sample municipalities were then selected from within each sample province with probability proportional to population size, without replacement. An additional provision was that each province must receive at least one municipality.
Stage 3. Allocation and Selection of Sample Spots
Once the sample provinces have been selected, 60 spots for each of the major areas were allocated among the sample provinces. Using the target number set for each spot in each region, the spots were distributed in such a way the each province was assigned a number of spots roughly proportional to its population size.
LUZON VISAYAS MINDANAO
Region I 8 Region VI 24 Region IX 9
CAR+REG II 8 Region VII 22 Reg X 9 Region III 14 Region VIII 14 CARAGA 7
Region IV 21 Region XI 17
Region V 9 Region XII 9
ARMM 9
---- ---- ----
TOTAL 60 60 60
Sample barangays within each sample municipality were selected with probability proportional to size.
Sample barangays were then classified as urban or rural based on the latest National Statistics Office classification (2000).
Stage 4. Selection of Sample Households
For the fourth stage, within each sample spot, five households were established by systematic sampling. In urban barangays as well as in rural barangays, designated starting points were randomly assigned - it was either: 1) a municipal/barangay hall, 2) a school, 3) the barangay captain's house, or 4) a church/chapel/mosque. A random start from 1-6 was also randomly generated for each spot. Thus, if a particular spot has a random start of 4, the first sample household should be the 4th household from the designated starting point. The sampling interval for urban barangays was six, while for rural barangays it was two.
Stage 5. Selection of Sample Respondents
For the fifth and final stage, as discussed earlier, a respondent was randomly chosen from among the voting-age adults in each selected household using a probability respondent selection table. A respondent not contacted during the first attempt was visited for a second time. If the respondent remained unavailable, or in cases where there was no qualified probability respondent of a given gender, the interval sampling of households would continue until five sample respondents were identified.
5. Research Methodology