Variable for Age-Specific Fertility Rate [message #12447] |
Thu, 18 May 2017 13:41 |
|
Dear Statalisters
I want to analyse potential sources of the differences in total fertility rates (TFRs) between two DHS surveys for Zimbabwe, 2005 and 2010. I aim to apply the decomposition method by Oaxaca-Blinder to identify contributions to fertility change.
I had initially planned to conduct the decomposition using the mvdcmp programme by Powers et al. (2011) using as my outcome variable children ever born (v201). The problem with v201 however, is that it is not age standardised and therefore not an optimal indicator for total fertility rate. In fact, average children ever born is higher in the Zimbabwe DHS2005-06 compared to ZDHS2010-11 whereas the ASFR and TFR estimates are higher in 2010-11 compared to 2005-06 survey. My intention now is to use age-specific fertility rate (ASFR) as a proxy for TFR as my dependent variable given that it is age standardised.
My question then is that how can I generate in STATA a variable for ASFR for single-year age groups (v012) in DHS. I tried the following;
For exposure [person-years];
gen top = v008 - 1
gen bot = v008 - 36
gen turn15 = v011 + 180
replace bot = turn15 if turn15 > bot
drop if bot>top
gen agebot = int( ((bot+top)/2 - v011)/12)
gen expo= top - bot + 1
Then for events [since I'm not well versed with looping];
gen birth1=0
replace birth1=1 if b3_01!=.
gen birth2=0
replace birth2=1 if b3_02!=.
gen birth3=0
replace birth3=1 if b3_03!=. I did this upto birth20 then generated variable for births by summing birth1 to birth20. However, deriving an ASFR indicator from here is not working out well.
I am aware of tfr2 which computes single-year age-group ASFRs through the tabexp command. My problem is that I want to generate a variable in the dataset which assigns by age (v012) an estimate of the ASFR that I will be able to use as an outcome variable in decomposition analysis.
I thank you in advance for your assistance.
Best regards,
Pedzisai Ndagurwa
[Updated on: Thu, 18 May 2017 13:44] Report message to a moderator
|
|
|
|
|
|
Re: Variable for Age-Specific Fertility Rate [message #18907 is a reply to message #18906] |
Fri, 13 March 2020 04:17 |
schoumaker
Messages: 66 Registered: May 2013 Location: Belgium
|
Senior Member |
|
|
Hello,
tfr2 allows you to estimate the effects of individual level variables on fertility.
For instance,
xi: tfr2 i.v106
will provide rate ratios by level of education.
xi: tfr2 i.v106 i.v025
will provide rate ratios by level of education and place of residence.
tfr2 uses a Poisson regression. These models assume that the age pattern of fertility is the same across categories of explanatory variables. This is usually OK, but you may want to relax that assumption.
The tabexp stata command (that comes with tfr2) can be used to prepare data that you cana analyze with a Poisson regression (or another model). tabexp will give you the number of births and exposure for the combinations of variables that you would like to include in your model.
For instance
xi: tabexp i.v106 i.v025
If you wan to take account of clustering in your model, type
xi: tabexp i.v106 i.v025, cluster(v001)
Hope this helps,
Best.
Bruno
Bruno Schoumaker
Centre for Demographic Research
Université catholique de Louvain
|
|
|
Re: Variable for Age-Specific Fertility Rate [message #18930 is a reply to message #18907] |
Sat, 21 March 2020 12:43 |
Jayanta
Messages: 28 Registered: September 2017
|
Member |
|
|
Dear Bruno,
Thank you for your prompt reply.
I've pooled seven rounds of DHS data. In this case, I would like to use the multi-level regression model with TFR (as dependent variable) and other socioeconomic variables (independent variables) such as education, wealth index, type of residence, religion, community-level variables (drawn from the person level response), etc. Also, how we can use the combined data to measure period wise TFR.
Please see the attached file, based on all seven rounds of data, which is the output of your given syntax. I can understand that this calculation gives a cluster wise calculation, but I wonder if I should use it for pooled data because it gives only one central estimate of the date. Also, how to coordinate this estimation data for regression model because there is no id available and several similar clusters. And the figures are based on a range of just 5 years(15,20,25,30,35,40,45).
Please help me in this regard.
Thank you in advance.
Regards,
Jayanta
|
|
|