I am working with birth weights in DHS 2000, 2004 and 2010 datasets for Malawi. Looking at the m19 variable (birth weight) the questionnaire appears to suggest that birth weight is recorded to 2 decimal places. However, when I look at the data the great majority of the birth weights only have 1 decimal place. For those with 2 or 3 decimal places, they occur in single frequencies. It is also surprising that the birth weight observations appear to cluster around weights that are factors of 0.5kg i.e. 1, 1.5, 2, 2.5 and so on

My question, was the data collected to two decimal places or was it deliberately rounded off to 1 decimal place. Second is the clustering of birth weights around factors of 0.5kg natural or is it due to some data collection/coding/cleaning protocol.

Thanks]]>

Quote:

M19, the birth weight of children is stored in the recode data file as weight in kilograms with three implied decimals or weight in grams. Implied decimals means that to use the variable in the unit indicated in the data description file or dictionary, you need to divide the number by the number of implied decimals.]]>

The country may elect to collect data a number of decimals other than three decimals. However, in the distributed data file, which is the recode data file, the value collected is converted into three decimals to make it comparable across surveys.

In all the surveys conducted in Malawi, the birth weight is collected either in Kilograms or in grams. No rounding was made for this value. The only explanation is that the clustering was done by the respondent when she recalled the birth weight of her child.