Child Discipline - Myanmar DHS 2015-2016 [message #22945] |
Wed, 09 June 2021 14:47 |
Nyan Linn
Messages: 8 Registered: December 2020
|
Member |
|
|
Dear Sir, May I ask some questions for my data analysis.
1. Which variable should be used to identify the primary caregiver of the child selected for child discipline module? sh15a or sh149? Are those two variables same or different? Could you please explain me?
2. Using sampling weight
I want to investigate the factors predicting violent disciplinary practices towards the children in Myanmar using multiple logistic regression (binary logistic regression).
Unit of analysis is child selected for child discipline module.
I restrict cases sh148==hvidx .
Directly using sampling weight of PR file is correct or wrong. (For both descriptive statistics and for using in svy commands)
If wrong, how should I modify the weight?
(I found some suggestions to multiply the weight by number of children 2-14 years in the household)
Is it really needed to do?
If I need to do like that, could you please instruct me to get the number of children aged 2 to 14 years in a household.
|
|
|
|
|
Re: Child Discipline - Myanmar DHS 2015-2016 [message #22952 is a reply to message #22948] |
Thu, 10 June 2021 12:14 |
Nyan Linn
Messages: 8 Registered: December 2020
|
Member |
|
|
Dear Sir,
Thank you very much for your instructions.
May I ask some more questions, please? Because I have some unclear parts.
My thesis title is family and household characteristics and violent disciplinary practices towards children in Myanmar.
The unit of analysis will be children aged 2- 14 years selected for the child disciplinary module. Is it correct?
For the descriptive result of my thesis, I will have descriptive tables of individual and sociodemographic characteristics of the children, and the prevalence of different forms of disciplinary practices towards the children.
I want to reproduce Table 17.5 Child discipline and Table 17.6 Child discipline by background characteristics of Myanmar DHS report book (in my thesis).
Could you please guide me to reproduce Table 17.6 Child discipline by background characteristics of Myanmar DHS report book?
I found your instruction in this user forum to reproduce Table 17.5 as follows.
ta sh11a
keep if sh11a==1
keep if hv102==1
keep if sh148==hvidx
gen wt=hv005/1000000
ta sh150 [iw=wt]
ta sh151 [iw=wt]
ta sh152 [iw=wt]
ta sh153 [iw=wt]
ta sh154 [iw=wt]
ta sh155 [iw=wt]
ta sh156 [iw=wt]
ta sh157 [iw=wt]
ta sh158 [iw=wt]
ta sh159 [iw=wt]
ta sh160 [iw=wt]
gen discipline = 0
replace discipline=1 if sh150==1|sh151==1|sh152==1|sh153==1|sh154==1|sh155==1|sh156= =1|sh157==1|sh158==1|sh159==1|sh160==1
ta discipline [iw=wt]
In the above instruction, you used gen wt=hv005/1000000.
But, in your instruction here, you make child weight .
egen neligible_children=total(sh11a), by(hv001 hv002)
gen child_wt=hv005*neligible_children if sh148==hvidx
So, if I use child_wt, the weight will be gen wt=child_wt/1000000. Right? If I use this weight, the weighted sample will be different from table 17.5 and table 17.6, right?
Could you explain to me why there are 2 weights and suggest to me which weight is more appropriate for my thesis?
Which weight should be used for my descriptive statistics?
Which weight should be used for svy command of my inferential statistics in simple logistic regression and binary logistic regression to find the predictors of different disciplinary practices?
Thank you very much, Sir, for explaining to me in detail.
With Deep Respect
Nyan Linn.
|
|
|
Re: Child Discipline - Myanmar DHS 2015-2016 [message #22992 is a reply to message #22952] |
Sun, 20 June 2021 20:51 |
Bridgette-DHS
Messages: 3208 Registered: February 2013
|
Senior Member |
|
|
Following is a response by Senior DHS Staff, Trevor Croft and Tom Pullum:
The child discipline module was developed at UNICEF for MICS surveys. The MICS tabulation plan says the following: "The child discipline module is administered in both the Questionnaire for Children Under Five and the Questionnaire for Children Age 5-17. The module is administered for all children age 1-4, but only to one randomly selected child age 5-17 (if age 5-14). To account for the random selection, the household sample weight is multiplied by the total number of children age 5-17 in each household; this weight is used when producing data on 5-14 year old children in this table." The references to the age range are inconsistent, but in the Myanmar survey the age range was 5-14.
DHS should have followed the same procedures, and weighted up the responses from the randomly selected child. If it did not do so, that was an error. We are too busy with other activities at this time to look further into this question.
|
|
|