Population Estimation Survey of Somalia 2014

Adjustments for non-response made on the data : Some EAs or settlements were not covered even though selected for the survey. Equally, some households listed in the EAs were not fully covered in a number of cases. There was need, therefore, to make adjustments for these two problems. In this case, considering the number of selected EAs to be ns and the number responding to be nr, the adjustment for the non-response involving EAs/settlements was ns/nr . Following the same approach, the adjustment for the households missed in the enumeration was given as nsh/nrh. These two adjustments were carried out on the sedentary urban and rural data. The number of the PSUs selected for the urban and rural is provided in Table 1 of the Methodology and Application Report, while the number of EAs that responded are provided in Table 2. The two tables provide a snapshot view of the extent of the non-response which consequently required adjustment.

Inaccessible areas were estimated independently using the satellite imageries. The satellite images were used to count the number of structures in the boundaries of each of the inaccessible areas. The information on mean sizes of structures, dwelling units and households were key in the application of the technique. From the listing information and the mapping exercise, the expected number of dwelling units per structure was established. In addition the average number of dwelling units in the structures was determined and using the mean household size of 6 the population for areas that could not be accessed was estimated. The use of this data was based on the fact that the averages for the variables under consideration do not vary widely particularly in the rural areas.

Corrections on data file : There were various duplicates in the data file involving records of individuals in the households. Within one household, there would be an individual having records duplicated. Such anomalies were detected by scrutiny of the records in the households. The members would be having same values for the variables of the data. These were cleaned out by deleting one of the duplicates whenever identified.

Nomadic data : The estimation was based on the weights based on the number of water points allocated to a given region and the mean number of watering episodes. Further adjustment was made in consideration of the duration of the watering episodes. One observation is that the mean watering time was not in conformity to the documented time for the animal types. For example it was noted that the camels were watering at closer intervals of time contrary to the 12 day period assumed in the study. This means that within the 12 day period there would be more camels appearing at the water points than expected. Thus if the watering interval was earlier understood to be 12 and yet in the interviewing period it is found to be 4 days, then there will be more camels at the water points during the interviewing period than expected in the theoretical formulation. Where such an observation was observed, an adjustment based on the ratio of the number of days n to 12 was used.