Findings, implications and lessons learned

  • As we have seen earlier for the two scenarios use of the second formula with ancillary data Mi in general reduced the standard error. This is again shown in the table below. All except the standard error for Lobamba were reduced when the second formula with Mi = number of homesteads with livestock was used.

    Use of ancillary data is useful if Mi is correlated with Ti.

     

    Dip-tank areas

    Standard errors

    Sub-region

    Population

    Sample

    Formula 1

    Formula 2

    Lobamba

    13

     4

    4,276(0.18)a

    5,327

    Malandzela

    12

     7

    4,310 (0.17)

    3,324

    Mayiwane

    20

    15

    6,960 (0.09)

    6,680

    Mbabane

    28

    15

    9,996 (0.12)

    6,883

    Ntfonjeni

    13

     7

    4,421 (0.13)

    3,347

    Piggs-Peak

    14

     5

    6,555 (0.14)

    4,481
             

    Total

    100

    58

     15,740

    12,756
    a ratio of standard error to population mean
  • Another point to note in the above table is how the standard error as a proportion of the population mean decreases as the sampling ratio of dip-tank areas increases. Contrast the ratio of 0.09 in the table above when 75% of dip-tank areas were sampled for Mayiwane Sub-region compared with the ratio of 0.18 when 31% of dip-tank areas were sampled in Lobamba Sub-region.