Home » Topics » Mortality » NMR, PMR, and IMR estimation with syncmrates
| NMR, PMR, and IMR estimation with syncmrates [message #13890] |
Fri, 19 January 2018 05:19  |
dgodha
Messages: 44
Registered: November 2016
Location: India
|
Member |
|
|
Hello Users,
I have been using the "syncmrates" command to estimate NMR, PMR, and IMR. The command used is as follows:
syncmrates v008 b3 b7 if v024==0 & v013==1 [iw=v005], cluster(v021)
For one of the countries, Stata returns this error message: 'r(cmr)' evaluated to missing in full sample"
In earlier datasets, CMR was simply omitted but I could still get estimates for the remaining mortality rates. I will appreciate if someone will suggest a solution.
Many thanks
Deepali
|
|
|
|
|
|
|
|
|
|
|
|
| Re: NMR, PMR, and IMR estimation with syncmrates [message #20061 is a reply to message #20051] |
Tue, 22 September 2020 10:46  |
Bridgette-DHS
Messages: 3214
Registered: February 2013
|
Senior Member |
|
|
Following is a response from DHS Research & Data Analysis Director, Tom Pullum:
At DHS we have heard about "syncmrates", a Stata package to calculate under-five mortality rates, but we have not actually used it. Perhaps some other forum users will respond.
Neonatal mortality is very easy to analyze with logit regression. In the BR file, in Stata, you define a variable such as "gen NNdeath=0" and "replace NNdeath=1 if b7==0". Then proceed with "logit NNdeath....", including covariates such as region on the right hand side. You will also want to include restrictions on the date of the birth (b3). If you convert the estimates back to probabilities, they will match with demographic approaches.
The other under-five rates are compound and not amenable to this approach.
|
|
|
|
| Re: NMR, PMR, and IMR estimation with syncmrates [message #20096 is a reply to message #20061] |
Fri, 25 September 2020 07:01 |
shujaat.smc@gmail.com
Messages: 75
Registered: July 2020
|
Senior Member |
|
|
Dear DHS Representative,
Thanks a lot for your response. I am analyzing the neonatal mortality determinants using survival analysis methodology.
I have use following variables to declare by data set as survival data;
stset b6 [iweight=v005], failure(childsurvivalstatus=1)
b6= age at death
childsurvival status= variable created from b5 (code=1 for death)
I have 31,683 missing value out of 34,425 (92% missing).
My Question1: Should I remove these missing values from my analysis (ie; drop these observation)
My Question2: In DHS statistics it is written that variable b7 (imputed age at death in months) is not allowed to miss, but the in my PDHS 2017-18 I have same amount of missing values as in b6 (age at death in days and months).
Waiting for your reply.
Best Regards
Dr. Hussain
stset b6 [iweight=v005], failure(childsurvivalstatus=1)
failure event: childsurvivalstatus == 1
obs. time interval: (0, b6]
exit on or before: failure
weight: [iweight=v005]
------------------------------------------------------------ ------------------
34425 total observations
31683 event time missing (b6>=.) PROBABLE ERROR
------------------------------------------------------------ ------------------
2742 observations remaining, representing
2742 failures in single-record/single-failure data
462674 total analysis time at risk and under observation
at risk from t = 0
earliest observed entry t = 0
last observed exit t = 326
|
|
|
|
| Re: NMR, PMR, and IMR estimation with syncmrates [message #20100 is a reply to message #20096] |
Fri, 25 September 2020 08:19 |
Bridgette-DHS
Messages: 3214
Registered: February 2013
|
Senior Member |
|
|
Following is a response from DHS Research & Data Analysis Director, Tom Pullum:
b6 and b7 are age at death and are only coded if the child has died (b5=0). If the child is still alive (b5=1), then b6 and b7 are not applicable (NA), which is coded with a dot in Stata. These are not "missing" values, they are just NA. Thus, your 92% "missing" actually means that 92% of children are living and 8% have died. In the terminology of survival analysis, age at death is censored for 92% of the children. I hope you are using the BR file for this analysis, rather than the KR file.
|
|
|
|
|
|
| Re: Neonatal Mortality Analysis [message #20142 is a reply to message #20120] |
Mon, 28 September 2020 14:33 |
Bridgette-DHS
Messages: 3214
Registered: February 2013
|
Senior Member |
|
|
Following is another response from DHS Research & Data Analysis Director, Tom Pullum:
The variable b6 has 3 digits. The first digit (counting from the left) is 1 if the units of measurement are days, 2 if months, 3 if years. This is mapped into b7, which is month. These are completed units. A child that dies in the first day has b6=100, for "units are days, and no days were completed" . That is, the death was on day 0. I believe that is the code you are looking for. I don't understand where "0.01" came from.
|
|
|
|
Goto Forum:
Current Time: Fri Dec 27 15:06:11 Coordinated Universal Time 2024
|
The DHS Program User Forum
Members
Search
Help
Register
Login
Home
