Versions causing changed sample size

[AMParticipant]

Why do the different version of ENA software like ver 2010 and ver 2011 give different sample size calculations in the survey planning stage even when the incidence, population proportion, prevalance, design effect etc. is the same? How will this effect the results of the survey if supposing the design has been planned in ENA 2010 but analysis is done in ver 2011?

[SMARTKeymaster]

The sample size calculation from different versions of ENA for SMART software should not change. However, the parameters for certain statistical tests in the plausibility report are modified from time to time to reflect the most recent findings which can result in slightly different plausibility report scores.

[AMParticipant]

Well from versions 2010 to 2011 the sample size variation is about 43 households.

[SMARTKeymaster]

There was a minor improvement made to the sample size calculation from 2010 to 2011 pertaining to the t-value and degrees of freedom when applied to cluster sampling and simple/random sampling.
ENA 2011 can be used to analyse data that used ENA 2010 for planning. Please note that a slightly larger sample size will result in a slightly more precise estimate. It is recommended to use the most recent versions of ENA for SMART for both planning and analysing of data.

[AMParticipant]

Sample Size calculation:
Estimated prevalence: 13.2%
Desired Precession: 5
Design effect: 1.7
Average Household size: 5.44
% of Under-5 Children: 12.68
% of non-responder household: 5
If we input all the information in ENA software 2010 and we got Sample size= 511 in ENA 2011 it is 552

Sample Size calculation:
Estimated prevalence: 11.8%
Desired Precession: 5
Design effect: 2
Average Household size: 5.70
% of Under-5 Children: 13.77
% of non-responder household: 10

If we input all the information in ENA 210 we got Sample size= 503 in ENA 2011 it is 548

These differences are large and not slight because they are almost a change of 10%.
If we have to explain to the client what is the justification to be given for the software to give a difference so large?

[SMARTKeymaster]

The justification is that ENA 2011 (all versions) will have a larger sample size; therefore, the confidence interval for the GAM estimate, along with other indicators, will be more precise (narrow).

[Author]

Posts