Is there anything obviously wrong with this approach, or anything you would do differently?
Note: I'm using Stata 15.
Code:
. proportion gender_n, vce(cluster person) over(year) citype(wilson) percent Proportion estimation Number of obs = 15,134 F: gender_n = F M: gender_n = M U: gender_n = U 2007: year = 2007 2008: year = 2008 2009: year = 2009 2010: year = 2010 2011: year = 2011 2012: year = 2012 2013: year = 2013 (Std. Err. adjusted for 8,237 clusters in person) -------------------------------------------------------------- | Robust Wilson Over | Percent Std. Err. [95% Conf. Interval] -------------+------------------------------------------------ F | 2007 | 59.85 1.17 57.55 62.12 2008 | 65.17 1.00 63.18 67.10 2009 | 69.34 0.81 67.73 70.90 2010 | 71.12 0.96 69.20 72.96 2011 | 67.63 1.04 65.57 69.62 2012 | 71.09 0.96 69.16 72.94 2013 | 66.56 1.32 63.93 69.09 -------------+------------------------------------------------ M | 2007 | 16.55 0.88 14.89 18.36 2008 | 19.97 0.84 18.38 21.67 2009 | 23.90 0.75 22.46 25.40 2010 | 20.63 0.86 19.00 22.36 2011 | 24.71 0.95 22.88 26.62 2012 | 22.14 0.89 20.46 23.93 2013 | 26.81 1.24 24.46 29.30 -------------+------------------------------------------------ U | 2007 | 23.59 1.01 21.67 25.62 2008 | 14.86 0.74 13.46 16.38 2009 | 6.76 0.44 5.95 7.68 2010 | 8.26 0.58 7.19 9.47 2011 | 7.67 0.59 6.59 8.90 2012 | 6.77 0.53 5.80 7.89 2013 | 6.63 0.69 5.39 8.12 --------------------------------------------------------------
0 Response to getting 95% CIs for multinomial variable, with clustering
Post a Comment