I assume that by wide area survey you mean the likelihood survey to continue with the Bayesian analysis? The purpose of the likelihood survey is to be able to use the data to strengthen your prior and come up with an overall posterior estimate of coverage. The answer to your question depends on the value or the meaning you ascribe to this overall estimate of coverage. If the estimating an overall coverage is your only aim (for many reasons one of which maybe is a need to report a single figure to a funder) regardless of spatial distribution of coverage, then I guess you might end up deciding to do a wide area survey. However, you need to be clear that the overall estimate is only meaningful if there is a strong indication that coverage across the area you are surveying is homogenous i.e. equally spatially distributed. If this is no the case, then the overall estimate is meaningless. The best example to illustrate this is the following figure from one of Mark's presentations on coverage which you can view [url=https://www.dropbox.com/s/cclksm7qdq4oqof/coverageDistribution.pdf]here[/url]. In the figure, you see that the different coverage estimates in each of the 8 sub-areas is 90% in 4 sub-areas and 10% in 4 sub-areas. The overall estimate is 50% across all 5 sub-areas but it is obvious that this estimate cannot be found anywhere in each of the sub-areas. This is what we mean by a meaningless overall estimate in the background of heterogenous spatial coverage. And in this context, I do not think proceeding to a likelihood survey to reach an overall estimate is meaningful but at the same time not practical and can even be misleading. A 50% overall estimate might make you think that you are doing well but in fact, there are 4 areas within your programme area that are failing significantly (10% coverage). From how you shared the data you were able to collect, it wasn't clear how you used the data to confirm or deny the hypothesis you made. Do you mind clarifying? But based on your last sentence, it seems that you have a strong indication that coverage is patchy. If this is the case, then as described in the example above, it seems that it is meaningless to estimate an overall coverage figure. More importantly, from a programme management perspective, an overall estimate figure may not necessarily give you more or valuable information aside from what you already have now that will help you reform your programme. Even without going for a likelihood survey to estimate an overall coverage figure, you have a rich set of information from stage 1 and stage 2 of SQUEAC that will guide you in reforming your programme. The added expense of doing a likelihood survey is not justified in this regard.

Ernest just answered your question and, up to a certain extent, you already had the answer: a coverage estimate in a patchy area will not give you any meaningful information. My point is another: you talk about active and adaptive case finding for SFP: for SFP, we usually do door to door, ACF is for SAM children; which method did you use?

I agree with what has been written here already. First ... a bit of history ... the wide-area survey is a recent addition to SQUEAC. It was added mostly because implementing agencies asked for it because donors demanded it. A wide-area estimate gives us a measure of the average coverage over the survey area. This may not reflect program experience. If (e.g.) coverage is 10% for one half of the population and 80% for the other half of the population then a wide-area estimate will be something like 45%. This level of coverage is not experienced anywhere in the program area. In this example, the average hides a lot of variation and this can mislead us. The example program (i.e. half at 10% and half at 80%) is very different from a program with a uniform (and experienced) 45% coverage. These two programs will need very different sets of reforms in order to improve coverage. You will usually be able to identify appropriate reforms from SQUEAC findings without doing a wide-area survey. Average coverage is "abstract" (i.e. not as experienced) when coverage is patchy but it does have some utility. It can be used to make compare different programs. It may be useful at the regional or national level to compare (map) coverage of district level programs and identify failing (or succeeding) programs. Using SQUEAC to do this at a national level, and quickly enough to allow meaningful comparisons, is likely to be rather expensive. We have SLEAC to do this sort of think quickly and cheaply. I hope this is of some use.