The DHS Program User Forum
Discussions regarding The DHS Program data and results
Home » Data » Weighting data » Confusion re: "Numbers in parentheses are based on 25-49 unweighted cases"
Confusion re: "Numbers in parentheses are based on 25-49 unweighted cases" Thu, 04 September 2014 14:14
 ABLR Messages: 10Registered: August 2014 Member
Greetings,

I am confused about a footnote that often appears in tables in DHS final reports, which reads "Numbers in parentheses are based on 25-49 unweighted cases". On the surface, it makes sense. However, I keep discovering what appear to be inconsistent applications of this "rule" in various final reports, so I suspect I am not understanding something.

For example, in table 10.9 on page 156 of the Bangladesh Final Report 2011 (screenshot attached), some estimates in the first line of the table (age in months: <6) are in parentheses and the number of children in the far right column is 51. However, further down the table in the "Division: Barisal" line, the total N in the far right column = 33, but none of the estimates are in parentheses. This is also true for Rangpur, Sylhet, and other divisions. I have discovered similar apparent inconsistencies in other final reports, so I don't believe this is an error.

The fact that the N for <6 months is over 50 (N = 51) but some of the estimates are in parentheses anyway suggests that this "rule" is based upon individual cell sizes (Numerator) rather than upon the total N used to calculate the estimate (denominator). That is, in this table, the number of children <6 months with ARI symptoms who were brought to a pharmacy, rather than the total number children <6 months with ARI symptoms (N = 51). However, if this were true, then
(A) all of the estimates for Barisal would also be in parentheses, because if the total number of children in Barisal is less than 50 (N = 33), then all of the individual cell sizes for each provider type would be <= 33
(B) many of the estimates for individuals with other characteristics (in the other divisions, mothers with complete primary) with Ns somewhat greater than 50, but not by much, would also be in parentheses.

Since they are NOT in parentheses, then I cannot figure out how the rule is applied.

In a nutshell, why aren't many other estimates in this table also in parentheses? And is this "rule" based on the number of unweighted cases in a given cell or based on the total number of cases in the denominator?

Re: Confusion re: "Numbers in parentheses are based on 25-49 unweighted cases" [message #2879 is a reply to message #2873] Fri, 05 September 2014 14:42
 Bridgette-DHS Messages: 2742Registered: February 2013 Senior Member
Following is a response from Senior Data processing Specialist, Ladys Ortiz:

I'm assuming that you are confused, because you don't quite understand the concept of "weighted" and "unweighted" cases. The Guide to DHS Statistics, states that: "Sampling weights are adjustment factors applied to each case in tabulations to adjust for differences in probability of selection and interview between cases in a sample, either due to design or happenstance".

The table you are analyzing is definitely correct. The estimate for "age in months: <6" is in parentheses because it's based on 48 cases UNWEIGHTED/51 cases weighted. The estimate for the "Division: Barisal", is not in parentheses because it is based on 63 cases UNWEIGHTED/33 weighted.

Please note that we do not generally include unweighted cases in the DHS reports.

[Updated on: Fri, 05 September 2014 14:42]

Report message to a moderator

Re: Confusion re: "Numbers in parentheses are based on 25-49 unweighted cases" [message #2880 is a reply to message #2879] Fri, 05 September 2014 14:51
 ABLR Messages: 10Registered: August 2014 Member
Thanks for your response. I do understand the weighting concept. However, since some of the UNweighted Ns I generated with the raw data match the Ns in the right-hand column (I didn't check them all), I had assumed that the Ns in the right column were unweighted. It appears, however, that this was just a coincidence.

[Updated on: Fri, 05 September 2014 14:53]

Report message to a moderator

 Previous Topic: Svysey vs iweght Next Topic: Ghana 2008 Birth Recode Weights
Goto Forum:

Current Time: Mon Mar 27 20:05:17 Coordinated Universal Time 2023