Dear all
I collected anthropometric data for my PhD, that I am analysing in STATA. I used the WHO macro for STATA to calculate Z-scores (WAZ, HAZ and WHZ) from my weight, height, age, and sex data for each child in my database.
Now that I am analysing my anthropometric outcomes, I have to go back to check a Z-score every now and then. Whan I have to manually calculate a Z-score, I do that doing ENA, because it's quicker than rerunning the STATA macro.
But what I find is that I get a different Z-score in WHO/STATA and ENA. Sometimes it’s a difference of almost 0.5 Z-score. On a similar note, in the field I used Z (Android app) to calculate Z scores for excluding kids WHZ <-3. Now, I’m finding some WHZs <-3 in the database – so perhaps Z and WHO/STATA give different results too.
Has anyone else ever ever found this discrepancy? How did you resolve this?
Many thanks.
bindi
Dear Bindi,
We had a similar discussion on this topic some years back, which you can find here
However, this only discussed the discrepancy between WHO and ENA, so perhaps others can shed light on the Z android app.
Best wishes,
Tamsin
Answered:
6 years agoDear Bindi:
Although perhaps not entirely germane, I recently calculated z-scores for all three indices (HAZ, WHZ, and WAZ) using the SPSS macros supplied by WHO. I then compared these z-scores to those derived from ENA. They were almost identical, with >99% of z-scores being the same to 2 decimal places and the remaining only differing by 0.01 z-score. So at least the SPSS z-scores are consistent with ENA.
Answered:
6 years agoHi Bindi,
The degree of accuracy and agreement of different tools to calculate z-score will depend on how accurate they handle age and how they allocate the LMS values to every value entered for calculation. Even within Stata, small differences will emerge, specially in extreme values, when comparing results obtained using the WHO macro, or the zanthro or zscore06 Stata commands.
Have you checked that when you are comparing values, you have entered the sex information correctly? Z-scores changes of the magnitude of about half standard deviation are commonly seen when incorrectly entering the sex of the child in the calculation.
Personally, I never use the WHO macro for Stata, as I find very clumsy, limiting and unfriendly. I usually calculate z-scores using zanthro command, as it allows for more options.
I have used the Z-score app in android phones, but I would advise against it as, in my experience, the app rounds up age value to the nearest month, affecting in turn the estimation of z-scores.
I hope the above is helpful
Answered:
6 years ago