Menu ENN Search
Language: English Français

SMART HHs Sampling

This question was posted the Assessment forum area and has 10 replies. You can also reply via email – be sure to leave the subject unchanged.

» Post a reply

Sameh Al-Awlaqi

Public Health and Nutrition Consultant

Normal user

23 Oct 2015, 17:05

Dear Colleagues,

We are planning to conduct a SMART survey in one area in Sudan. This area is a mixture of IDPs camps and host communities with no specific arrangement of HHs or streets. The estimated population is around 102,000.There is no accurate mapping and it will take time to count all HHs within each of the 34 clusters if we wanted to use systematic random sampling method.
What is the best method to select segments and HHs in this case? My colleagues from Ministry of health use " the spinning pen" technique in SMART surveys by going to the center of the cluster, spin a pen, go with the direction of that pen to the end of the cluster, spin it again, record all HHs on their left and right, then randomly select one as a start to interview and continue until finishing their target households.They are saying it is a good one for their context.

I know this method has been used in the past for some nutrition assessments but I am not sure if it is the right one to be used here in SMART given the challenges mentioned above.

Thanks!

Sameh

Anonymous 81

Public Health Nutritionist

Normal user

25 Oct 2015, 18:33

Dear Sameh,
I can see two questions from your post. The first is how to deal if the survey population is mixture of IDPS and host communities. Are you assuming both are the same condition? As you stated, it seems there is no clear map of the villages or enumeration areas. This issue is everywhere. If you managed stage I (listing of geographic area and selection of clusters), in the next stage, the villagers can assist you to locate their own catchment.
Regarding the selection of HH method at stage II, given the SMART guideline, I would advise you to use either random or systematic sampling method. Now days, modified EPI method is not advisable. What you need is just listing households regardless of HH size or whether or not there are children under five. Then you select the required HHs using either random or systematic

Sameh Al-Awlaqi

Public Health and Nutrition Consultant

Normal user

26 Oct 2015, 07:02

Hi Kiross,

Thanks for you reply.

We have managed to select the clusters in stage one, they have similar conditions. If modified EPI method is not advisable, could implementing it put question marks on the reliability of findings from the survey even if we have higher precision and more HHs/children to be interviewed and screened?

Regards,Sameh

Anonymous 81

Public Health Nutritionist

Normal user

26 Oct 2015, 08:25

Dear Sameh,
I am coping the whole statement shown below from one of the SMART guide (June 2012). as follow

This method is the least recommended one because it induces certain biases in the
sampling:
1. Since we always select the closest household on the right of the previous one, or
the one that has the closest door, it creates a bias because the selection of every
household is not independent from the selection of the previous one (and this is
one important criterion for probability sampling).
2. Also, by choosing households by proximity, you are more likely to stay within the
same area of the village, while households close to each other are more likely to
have the same characteristics, and hence, you might not be getting a
representative sample of your cluster.

Mark Myatt

Consultant Epidemiologist, Brixton Health

Frequent user

26 Oct 2015, 12:14

The "spin the bottle" EPI-derived technique seems to work OK for anthropometry surveys. It does not work well for some indicators. There is a tendency for loss of sampling variation leading to high design effects. You can fix this (to some extent) by having more smaller clusters. The issue is loss of sampling variation not bias.

There are alternative within-cluster sampling methods. I find the newer SMART approaches rather impractical. Modification of the EPI method (EPI3 and EPI5) in which the EPI method is used to sample the first HH with subsequent HHs selected by spinning the bottle again and taking the third (or fifth) HH in the indicated direction works well for many indicators. Such methods assume a simple community structure.

You may want to use the within-cluster sampling methods used in RAM and RAM-OP type surveys. There is a manual for RAM-OP here. This method can sample from simple and complicated communities.

I hope this helps.

Sameh Al-Awlaqi

Public Health and Nutrition Consultant

Normal user

26 Oct 2015, 12:39

Many thanks Mark for you reply,

I will read the the RAM-OP manual and will revert back should I have any questions.

Thanks again,

Regards,Sameh

Bradley A. Woodruff

Self-employed

Technical expert

26 Oct 2015, 18:05

In fact, there are many ways to randomly select an unbiased sample of sampling units, but the EPI "spin-the-pen" technique is not one of them. One of the presentations in the original SMART workshop in 2002 explained why (see http://web.archive.org/web/20081119155245/http://www.smartindicators.org/workshop/a/woodruff-pm-23.htm).

However, nothing is perfect, and sampling, like life itself, requires compromise between correctness and expediency. My preference in a specific situation is to initially consider the most correct sampling technique possible and be willing to expend some energy and resources to use this method. The most statistically and epidemiologically correct method is listing all sampling units and using simple random selection to choose among them. If this is impossible or logistically too demanding, select a feasible, albeit perhaps less correct, method with the full understanding of potential effects on sampling error and bias. The really important point is that, regardless of SMART or any other standard recommendations, you must choose the most appropriate sampling strategy for a specific situation, and this requires knowledge and experience. No cookbook can tell you how to sample in every situation.

There are several sources which describe the process of selecting the most correct and feasible sampling method:
1. A Manual: Measuring and Interpreting Malnutrition and Mortality (see: http://www.unhcr.org/45f6abc92.pdf)
2. Gary T. Henry. Practical Sampling. SAGE Publications, Applied Social Reseach Methods Series Volume 21. 1990.
3. The presentation at the SMART workshop cited above.

Kemal J Tunne

Mr

Normal user

26 Oct 2015, 18:33

Thanks every one I learn a lot from the manual RAM-OP its better to follow the guidelines for perfect sampling

Anonymous 310

Normal user

7 Nov 2015, 09:58

As others pointed out, the EPI method is not advisable for sampling. If the ‘clusters’ are already selected (assuming at village level), then the list of the HHs at the sampled cluster often done using the villagers and local knowledge in most context in south sudan (the survey team doesn’t necessary need to move around the entire village to list the HHs). So, active involvement of the local people at the ground during the planning stage is very important. If the selected cluster (village), is large or the HHs dispersed as you described, then 'segmentation' can be done (assuming the village has >than 100 HHs). At the end, you will end up of using either random or systematic sampling to select the required HHs in the sampled cluster. I understand the challenges in some circumstances, but we should do our utmost to avoid sampling error/bias during the planning stage.

RAM-OP method applies different methodology for selection of clusters/PSUs at Stage One (not PPS)

Anonymous 730

Normal user

7 Nov 2015, 10:01

Spot on!

Kennedy Musumba

SMART Program Manager

Normal user

17 Nov 2015, 11:41

During the second stage sampling, it is advisable to liaise with the village leaders/authorities to come up with an updated list of households per cluster (sampling frame) dependent of your household definition and subsequently apply simple random or systematic random sampling.

In cases where clusters are large (approximate more than 100 households), Segmentation is applied based on PPS followed by either simple random or systematic random sampling. Further detailed information can be obtained from;
1.Module 3-Sampling
2. Sampling for SMART, Pgs. 26-32 (Complementary Tools &Resources, download Handouts)
http://smartmethodology.org/survey-planning-tools/smart-capacity-building-toolbox/

It is important to note that EPI method /pen spinning is not as highly recommended because of its limitation of introducing selection bias.

Back to top

» Post a reply