Menu ENN Search
Language: English Français

interpreting weight for height standard deviation from ENA plausibility check

This question was posted the Assessment forum area and has 2 replies. You can also reply via email – be sure to leave the subject unchanged.

» Post a reply

Anonymous 1740

Normal user

19 Oct 2015, 18:01

When conducting a SMART survey; after running the plausibility check using ENA you realise that the weight for height standard deviation of some teams' data is close to out of the range of between 0.8 and 1.2 (the lowest is 0.81 and the highest is 1.11, current overall standard deviation is 0.89). What could be the possible causes and how can this be addressed. Data collection is still ongoing.

Damien Pereyra

Independant Nutrition Evaluation Expert

Normal user

20 Oct 2015, 08:37

Hello,

First of all, in a survey you have to consider the overall data set (=sample), if there is many strata, each strata=1 sample. However, it is important during data collection to manage daily analyses of plausibility check for each surveyor team. Could you share in this discussion the first table of this repport for a better understanding.
For ex:
Overall data quality
Criteria Flags* Unit Excel. Good Accept Problematic Score
Flagged data Incl % 0-2.5 >2.5-5.0 >5.0-7.5 >7.5
(% of out of range subjects) 0 5 10 20 0 (2,4 %)
Overall Sex ratio Incl p >0.1 >0.05 >0.001 <=0.001
(Significant chi square) 0 2 4 10 0 (p=0,212)
Age ratio(6-29 vs 30-59) Incl p >0.1 >0.05 >0.001 <=0.001
(Significant chi square) 0 2 4 10 0 (p=0,144)
Dig pref score - weight Incl # 0-7 8-12 13-20 > 20
0 2 4 10 2 (8)
Dig pref score - height Incl # 0-7 8-12 13-20 > 20
0 2 4 10 0 (4)
Dig pref score - MUAC Incl # 0-7 8-12 13-20 > 20
0 2 4 10 0 (7)
Standard Dev WHZ Excl SD <1.1 <1.15 <1.20 >=1.20
. and and and or
. Excl SD >0.9 >0.85 >0.80 <=0.80
0 5 10 20 0 (1,06)
Skewness WHZ Excl # <±0.2 <±0.4 <±0.6 >=±0.6
0 1 3 5 0 (-0,07)
Kurtosis WHZ Excl # <±0.2 <±0.4 <±0.6 >=±0.6
0 1 3 5 1 (-0,26)
Poisson dist WHZ-2 Excl p >0.05 >0.01 >0.001 <=0.001
0 1 3 5 0 (p=0,390)
OVERALL SCORE WHZ = 0-9 10-14 15-24 >25 3 %
The overall score of this survey is 3 %, this is excellent.

Or send you plausability chech repport in bilateral by email: damienpereyra@gmail.com
Thanks
Thanks

Mark Myatt

Consultant Epideomiologist

Frequent user

20 Oct 2015, 09:07

Checking like this as you enter data can be difficult. It is possible that the team with the highest SD is measuring with less precision than the other teams. It is also possible that the teams with lower SDs are submitting data made from a small set of clusters (i.e. they are making up "new" data from previously collected data - I have seen this several times). It is also possible that the team with the large SD have sampled from a less homogenous set of clusters than other teams. It is also possible that what your are seeing is due to chance (particularly as you will only have small samples in which to estimate the SD).

I think you need to look at the data carefully for (e.g.) copying (sorting on a set of variables can pick up duplicates), digit preferences, &c. (you should do this anyway) and increase supervision of the highest and lowest SD teams.

You may find that you have little to worry about.

BTW : I am a bit of a heretic when it comes to these checks. I see no good reason to assume perfect normality in WHZ data in all settings and no good reason to automatically reject survey data with SD > 1.2.

I hope this is of some use.

Back to top

» Post a reply