I am going to do a study on ANC service utilization in urban slums of Kathmandu. My sample size on the basis of this formula: N= Z2PQ/L2 is 92 with 10% allowable error and 10% non-response rate.I am taking each slum as a cluster. I am going to select the cluster purposively and am planning to do the census in selected cluster as I couldn't get the total number of my study population in each slum .So, in this case, how do I calculate design effect to find my actual sample size.
Pramila -
A design effect compares the efficiency of a given more complex design to what you could have gotten with a simple random sample of the same sample size. In general one would expect a stratified random sample to do better, and a cluster (random) sample to do worse, but make up for it in logistical help, which makes a larger overall sample size more feasible.
However, you do not appear to be doing a cluster sample. It sounds like you are doing censuses of given "slums," each selected purposely. Do you have any reason to believe these 'slums' are representative of others? You do not mention any model that shows that they are. At any rate, this is not what is meant by cluster sampling. The clusters would have to have been chosen at random.
Best wishes - Jim
PS - I think you may be confusing some procedures that have specific, conflicting assumptions about the data and the data collection.
PSS - You might be interested in the following:
Kish, L.(1995), “Methods for Design Effects,” Journal of Official Statistics, Vol. 11, No. 1, pp. 55‐57.
http://www.jos.nu/Articles/abstract.asp?article=11155
(click on "Full Text")
- Badges
- Science topic
- Similar topics
- Publishing