percentile graph based on event time

[Zariab Hossain]

Dear Statausers

I have the following dataset, where I have municipality (kom_kod), redovar (year), number of office (office_WS), number of caseworkers working at offices (CW_WS), closeyear (closing year of municipality offices) and event time (0 means when the municipality office closes down):

[CODE]
* Example generated by -dataex-. For more info, type help dataex
clear
input int kom_kod float redovar double(office_WS CW_WS) int closeyear float event_time
115 2005 1 7 2015 -10
115 2006 1 12 2015 -9
115 2007 1 9 2015 -8
115 2008 1 3 2015 -7
115 2009 1 3 2015 -6
115 2010 1 4 2015 -5
115 2011 1 3 2015 -4
115 2012 1 4 2015 -3
115 2013 1 3 2015 -2
115 2014 1 3 2015 -1
115 2015 1 3 2015 0
115 2016 0 0 2015 1
115 2017 0 0 2015 2
115 2018 0 0 2015 3
117 2005 1 6 2016 -11
117 2006 1 8 2016 -10
117 2007 1 7 2016 -9
117 2008 1 7 2016 -8
117 2009 1 6 2016 -7
117 2010 1 11 2016 -6
117 2011 1 12 2016 -5
117 2012 1 23 2016 -4
117 2013 1 19 2016 -3
117 2014 1 16 2016 -2
117 2015 1 14 2016 -1
117 2016 1 1 2016 0
117 2017 0 0 2016 1
117 2018 0 0 2016 2
125 2005 1 3 2015 -10
125 2006 1 3 2015 -9
125 2007 1 2 2015 -8
125 2008 1 2 2015 -7
125 2009 1 2 2015 -6
125 2010 1 3 2015 -5
125 2011 0 0 2015 -4
125 2012 0 0 2015 -3
125 2013 1 1 2015 -2
125 2014 1 1 2015 -1
125 2015 1 2 2015 0
125 2016 0 0 2015 1
125 2017 0 0 2015 2
125 2018 0 0 2015 3
139 2005 2 4 2015 -10
139 2006 2 4 2015 -9
139 2007 2 2 2015 -8
139 2008 0 0 2015 -7
139 2009 0 0 2015 -6
139 2010 0 0 2015 -5
139 2011 0 0 2015 -4
139 2012 0 0 2015 -3
139 2013 0 0 2015 -2
139 2014 0 0 2015 -1
139 2015 0 0 2015 0
139 2016 0 0 2015 1
139 2017 0 0 2015 2
139 2018 0 0 2015 3
192 2005 1 5 2017 -12
192 2006 1 6 2017 -11
192 2007 1 8 2017 -10
192 2008 1 2 2017 -9
192 2009 1 2 2017 -8
192 2010 1 6 2017 -7
192 2011 1 4 2017 -6
192 2012 1 2 2017 -5
192 2013 1 3 2017 -4
192 2014 1 3 2017 -3
192 2015 1 10 2017 -2
192 2016 1 6 2017 -1
192 2017 1 5 2017 0
192 2018 1 4 2017 1
305 2005 1 2 2016 -11
305 2006 1 2 2016 -10
305 2007 1 3 2016 -9
305 2008 1 4 2016 -8
305 2009 1 5 2016 -7
305 2010 0 0 2016 -6
305 2011 1 4 2016 -5
305 2012 1 5 2016 -4
305 2013 1 4 2016 -3
305 2014 1 3 2016 -2
305 2015 1 9 2016 -1
305 2016 1 4 2016 0
305 2017 1 1 2016 1
305 2018 0 0 2016 2



I have done the some event time figures using the following code:

keep if inrange(event_time,-5,5)
gcollapse (mean) CW_WS office_WS, by(event_time)
label var CW_WS "average CW and PES office"
*label var office_WS "average PES office"
grstyle init
grstyle set plain, dot horizontal grid

twoway (line CW_WS event_time, legend(off) yaxis(1) ) ///
, graphregion(color(white)) plotregion(color(white)) xline(0, lcolor(black) lpattern(dash)) graphregion(color(white)) name(a, replace) ///
title("Average CW per kommun-year", size(small)) subtitle("(for kommun closed between 2014-2017)", size(small))
graph save "$figures/CW_WS.gph", replace

twoway (line office_WS event_time, legend(off) yaxis(1)) ///
, graphregion(color(white)) plotregion(color(white)) xline(0, lcolor(black) lpattern(dash)) graphregion(color(white)) name(b, replace) ///
title("Average office_WS per kommun", size(small)) subtitle("(for kommun closed between 2014-2017)", size(small))
graph save "$figures/office_WS.gph", replace


grc1leg "$figures/CW_WS.gph" "$figures/office_WS.gph"
graph export "$figures/CW_and_PES_office_WS.pdf", replace




Now, I want to do this graph for 20th percentile group of the distribution. I am pretty unclear about how to do that. Any help would be highly appreciated.

Thanks,

Zariab Hossain
Uppsala University

[George Ford]

Maybe instead of mean collapse on p20.

or else use pctile to dummy them out and then "if" on collapse.

[Zariab Hossain]

Thanks George for the tip. I also did this and got what I wanted.