Statistical modelling/Milk consumption & marketing

Taking into account the sizes of the standard errors, and the different ways in which the data were collected, the results for Assa look reasonable, but those for Bilisa not. One thus needs to consider carefully the quality of data collected and possible sources of error.

The large proportion of households that did not provide information on quantities of milk consumed in the households in Bilisa places a question mark on the justification of using a mean from the remaining households. By dividing amounts of milk consumed per household by family size in the CS1Data2 edited spreadsheet it can be seen that there is a wide range in individual proportions and thus a poor correlation. One reason, both for this poor correlation and the lack of response in some cases, was that the head of the household often said that he could not give an answer because milking and milk management largely fell into the hands of his wives.

In view of the possible inaccuracies in the collection of these data it will be unwise to publish the results shown in the table for wider circulation.

The limited conclusions that one might draw are that households in Assa had very little milk to spare. For households in Bilisa, however, where average milk offtakes were higher, there are opportunities for marketing, but the precise amounts of milk available for this purpose are not clear.