Normal user
15 May 2012, 04:49
WE have faced a situation when using ENA 2011 November version. What happened is the result of index of dispersion on the results anthropometry page and plausibility test are different. What is the reason for this discrepancy (see a result from a survey for the two outputs below)? Another related question is that the plausibility test check the validity or quality of the survey based on the age data. This quality test is assuming a developing country with high mortality rate in the under five population (based on the default on the option page). But it is obvious that there are different scenarios of mortality for the under fives, including very small or no mortality, fast declining mortality and what not. The second question will be is there any default that account for this variability and make the plausibility check more usable? (Or is there any work by demographers to account for specific contexts?). Many think that if we get penalized for not conforming with the softwares default the survey quality are poor; which is not the case according to SMART training. Your help in this is highly appreciated.
Results of ID from ENA results page
WHZ < -2: ID=0.74 (p=0.883)
WHZ < -3: ID=0.92 (p=0.607)
GAM: ID=0.74 (p=0.883)
SAM: ID=0.92 (p=0.607)
HAZ < -2: ID=1.68 (p=0.005)
HAZ < -3: ID=0.98 (p=0.513)
WAZ < -2: ID=1.37 (p=0.063)
WAZ < -3: ID=1.15 (p=0.241)
Results of ID from the plausibility word page:
Test if cases are randomly distributed or aggregated over the clusters by calculation of the Index of Dispersion (ID) and comparison with the Poisson distribution for:
WHZ < -2: ID=0.86 (p=0.718)
WHZ < -3: ID=1.08 (p=0.342)
GAM: ID=0.86 (p=0.718)
SAM: ID=1.08 (p=0.342)
HAZ < -2: ID=1.39 (p=0.055)
HAZ < -3: ID=1.04 (p=0.404)
WAZ < -2: ID=1.11 (p=0.297)
WAZ < -3: ID=0.91 (p=0.638)
Frequent user
15 May 2012, 14:39
I passed this question to Oleg Bilukha at CDC. He replied:
He should check what growth standards he has selected on
the results page. If NCHS is selected, then there will be
a discrepancy in ID scores since plausibility check is
always done using WHO standards.
He added:
Please advise him to direct his further questions on
SMART (as well as anybody else’s questions related to
SMART posted on the ENN blog) to ACF CA team at the
smartmethodology.org website blog.
The forums are
here.
I think that we should ask ACF-CA to monitor this site so that can respond to questions posted here.
I hope this is of some use.
Normal user
15 May 2012, 20:29
Thanks Mark and Oleg. I have tried to post this question on the SMART community page and I am waiting approval to join the community from moderators.
Thanks
Normal user
15 May 2012, 20:32
I selected NCHS for the results and now I understood why the discrepancy happened. Thanks