Menu ENN Search
Language: English Français

SMART urban sampling

This question was posted the Assessment forum area and has 5 replies. You can also reply via email – be sure to leave the subject unchanged.

» Post a reply

Maja Kuhar

Normal user

5 Feb 2014, 17:43

We have been asked to do a SMART survey in a township of a large African city. The section is a mix of blocks with pockets of less ordered housing. We are trying to get the maps and population of enumeration areas.
How would you recommend to sample in this type of setting? What sample sizes do we need?

Anonymous 81

Public Health Nutritionist

Normal user

6 Feb 2014, 05:21

Dear Maja
To calculate the sample size, the same principle applies whether it is urban or rural. The sample size depends on the context of the following parameters; population, expected prevalence, precision and design effect. With regard to sampling methods or procedure in the second stage which is selecting of household, the feasible method is either use simple random or systematic sampling. To do this, you might need about one extra day for listing or updating of households in the respected enumeration area or cluster. .

Mark Myatt

Consultant Epideomiologist

Frequent user

6 Feb 2014, 08:22

I took a look at the SMART manual. It only mentions urban settings in the contexts of food security assessments and says "Assessment in ... urban areas requires some modification of technique and is not discussed in this manual" (page 90).

The sampling material (page 35 onwards) seems to suggest systematic or random sampling. I doubt you will have the lists required for random sampling but you could walk every street taking (e.g.) every twelfth (i.e. if the sampling interval is 12) doorway. To know the sampling interval you will need to count doors (all doors on all streets) and then find out (small survey) how many eligible children are, on average, behind each door. This is covered from page 50 of the SMART manual. The manual also suggests that a cluster sample might be used but is a little vague on this point.

There is also a paper on sampling methods on the SMART site. This has more detailed guidance although nothing specifically on urban settings.

I took a look at the SMART forum but it looks very quiet over there. The last post was a year ago. You might give this a try. I couldn't work out how to log-in and ask a question.

It isn't SMART but this manual from the WHO might prove useful.

I think you need to be very careful as some donors and governments are very strict about what they allow or accept as a "SMART survey" and, in my experience, the SMART people can get nasty when it comes to alternative methods. You need to be sure that what you do is what the SMART manual tells you to do or have someone from SMART give their approval.

I hope this helps.

Mark Myatt

Consultant Epideomiologist

Frequent user

6 Feb 2014, 08:33

I forgot to mention ...

UNICEF and the ministries of health in Sudan have been using spatial sampling methods in their CNS (RAM type surveys) and S3M surveys which include urban settings. VALID and GAIN have been using S3M samples with variable density to sample per-urban and urban settings in Ghana and India. This methods are definitely not SMART (they are intelligent but not "SMART") and so may not be acceptable to some donors and governments.

Mark Myatt

Consultant Epideomiologist

Frequent user

10 Feb 2014, 08:41

Can we have someone from SMART respond to this question?

Victoria Sauveplane

Senior Program Manager, Action Against Hunger CA

Normal user

11 Feb 2014, 13:58

Dear Maja,

Urban sampling is generally quite tricky to undertake. Depending on the context, some cities are easier to sample versus others. As Tefera has mentioned, your sample size should not change if your survey area was in a rural or urban setting. I would suggest spending an extra day before doing HH selection to enumerate the households (build a list if the area is small enough and there is no geometric order, aka rows of households) and apply simple random sampling with a random number table. Alternatively, if the households are arranged in rows, you can directly apply systematic random sampling and walk to the households that correspond to your sampling interval (Total number of households in a given area/number of households to survey in a cluster). In the case that the survey teams come across an apartment building, make sure to enumerate or count all of the households in that building to allow equal chance of all residents to be selected.

As explained in the SMART Sampling Paper [link: http://www.smartmethodology.org/documents/sampling/Sampling_Paper_June_2012.pdf], segmentation would be a good method to use to reduce a given neighborhood into more manageable blocks to perform household selection.

For all SMART-related questions, please refer to the SMART methodology website (www.smartmethodology.org) or you may also contact me directly.

Many thanks and best of luck,

Back to top

» Post a reply