| Re: Standardizing glucose measurements across surveys [message #28393 is a reply to message #28385] |
Tue, 26 December 2023 17:34  |
Bridgette-DHS
Messages: 3199
Registered: February 2013
|
Senior Member |
|
|
Following is a response from Senior DHS staff member, Tom Pullum:
You would answer a question such as this with the following steps.
First, find the variable and the label in the PR file. This took me a couple of tries, but I finally found it with "lookfor hba":
shwhba1c int %8.0g SHWHBA1C final result hba1c for women (3 decimals implicit)
The label of the variable suggests that the values include a factor of 1000, but I proceed to check this.
Second, list the label to find the non-numeric codes:
. label list SHWHBA1C
SHWHBA1C:
99995 inconclusive
This value must be excluded from any data analysis.
Third, summarize the numeric values of the variable:
. summarize shwhba1c if shwhba1c<99995
Variable | Obs Mean Std. Dev. Min Max
-------------+---------------------------------------------------------
shwhba1c | 4,331 6392.265 1344.853 3800 21400
Fourth, search the report to find the typical values in tables. I see that the cutoffs are 5.7% and 6.4%. Multiplied by 1000, these are 5700 and 6400. Those are in the range from 3800 to 21400, and with any other factor of 10 the values would be outside the range from 3800 to 21400.
Answer: the variable in the data file includes a factor of 1000.
|
|
|
|
The DHS Program User Forum
Members
Search
Help
Register
Login
Home
