Home » Data » Weighting data » Clustered Standard Errors
Clustered Standard Errors [message #9219] |
Thu, 25 February 2016 13:10 ![Go to next message Go to previous message](/theme/default/images/down.png) |
cbdolan
Messages: 17 Registered: March 2013 Location: Williamsburg, VA
|
Member |
|
|
I am using the 2007 and 2013/14 DRC DHS Birth Recode Files.
I have set up the weights and adjustments for DHS surveys using the following syntax:
gen wgt=v005/1000000
egen stratum=group(ADM1_CODE* v025)
svyset [pw=wgt],psu(v021)strata(stratum)
Am I correct that the svyset command, when called with svy: at the start of a regression, produces robust SE clustered at the cluster level? or are these robust standard errors? Is it possible to correctly use the svyset command and cluster the SE at the ADM1(province level)or is it better to not call the svy: command and do the following:
regress y a b c...cluster(ADM1_CODE)
*please note: in the 2007 DRC DHS the v024 variable which is typically used in DHS adjustments contains both numeric and character values for the same province (see below). Therefore, I used ADM1_CODE and not v024 when making unique strata values by region/urban-rural
province | Freq. Percent Cum.
-----------------+-----------------------------------
kinshasa | 106,141 4.86 4.86
bandundu | 231,557 10.60 15.47
bas-congo | 86,226 3.95 19.41
equateur | 251,785 11.53 30.95
kasai-occidental | 153,546 7.03 37.98
kasai-oriental | 187,879 8.60 46.58
katanga | 214,694 9.83 56.41
maniema | 89,496 4.10 60.51
nord-kivu | 104,015 4.76 65.28
orientale | 224,819 10.30 75.57
sud-kivu | 104,501 4.79 80.36
20 | 33,727 1.54 81.90
30 | 44,293 2.03 83.93
40 | 51,749 2.37 86.30
50 | 33,135 1.52 87.82
61 | 40,326 1.85 89.66
62 | 44,224 2.03 91.69
63 | 40,950 1.88 93.57
70 | 47,891 2.19 95.76
80 | 48,764 2.23 97.99
90 | 43,851 2.01 100.00
-----------------+-----------------------------------
Total | 2,183,569 100.00
|
|
|
Goto Forum:
Current Time: Fri Aug 30 21:25:40 Coordinated Universal Time 2024
|