Problem sketching hourly time labels on x (or t) axis of a two-way graph

[Alireza Mahdavi]

Hi experts;

I am trying to sketch a twoway rcap figure showing the start and end time of a few people's performances (horizontally);

Here are my xvar1 and xvar2 (showing the start and end points of rcap). My y-axis values are also the list of people performing during these time series.

Start_Time End_Time
08:30:00 16:00:00
12:00:00 13:30:00
12:00:00 13:00:00
13:30:00 15:15:00
13:30:00 15:29:59
14:00:00 15:00:00
14:00:00 15:00:00
14:30:00 15:29:59
14:45:00 15:00:00
15:00:00 16:59:59
15:00:00 16:00:00
15:30:00 16:14:59
16:00:00 16:45:00
16:30:00 17:30:00
17:00:00 18:00:00
17:00:00 18:00:00
17:00:00 18:00:00
17:30:00 18:29:59
17:30:00 19:30:00
18:00:00 20:15:00
18:30:00 19:30:00
19:00:00 21:00:00
19:30:00 21:00:00
20:00:00 21:45:00
20:30:00 21:40:00
21:00:00 22:30:00
21:00:00 21:29:59
22:00:00 22:59:59
23:00:00 01:59:59
23:30:00 01:59:59

My problem is that I cannot label the x (or t) axis (the horizontal axis) properly. I know to with the numbers (not hours) it is easy to create lables like below:
xscale (range(0 24))
xlabel(0 (2) 24) // showing 0 2 4 6 ...... 22 24 on your x-axis

But here with the time, I don't know how to do it like the numbers.

Here is what I want:
08:00:00 09:00:00 10:00:00 ....... 21:00:00 22:00:00 23:00:00 // or even better below
08:00 am 09:00 am 10:00 am ........ 08:00 pm 09:00 pm 10:00 pm

The format of Start_Time and End_Time above is %tcHH:MM:SS.

Can you please help me what I should do to label my figure x-axis like what I described?

Thanks

[Andrew Musau]

Please use dataex in the future to present data examples. You can use the clock() function to access the specific SIF time values, e.g.,

. di clock("00:00:00", "hms")
0

. di clock("01:00:00", "hms")
3600000

. di clock("23:59:59", "hms")
86399000

Once you have these, you proceed the same way as you would with numerical values, only specifying the format.

* Example generated by -dataex-. To install: ssc install dataex
clear
input float id double(starttime endtime)
 1 30600000 57600000
 2 43200000 48600000
 3 43200000 46800000
 4 48600000 54900000
 5 48600000 55799000
 6 50400000 54000000
 7 50400000 54000000
 8 52200000 55799000
 9 53100000 54000000
10 54000000 61199000
11 54000000 57600000
12 55800000 58499000
13 57600000 60300000
14 59400000 63000000
15 61200000 64800000
16 61200000 64800000
17 61200000 64800000
18 63000000 66599000
19 63000000 70200000
20 64800000 72900000
21 66600000 70200000
22 68400000 75600000
23 70200000 75600000
24 72000000 78300000
25 73800000 78000000
26 75600000 81000000
27 75600000 77399000
28 79200000 82799000
29 82800000  7199000
30 84600000  7199000
end
format %tcHH:MM:SS starttime
format %tcHH:MM:SS endtime

tw (rcap starttime endtime id, xlab("") ylab(3600000(3600000)86399000, ///
format(%tcHH:MM:SS) angle(0)) scheme(s1color) xtitle("") ytitle(Start/End ))

If you want 1:00 am, 2:00 am, etc. on the axis, you will have to use value labels for your time variable. There are plenty of examples in the forum on how to do this if you search.

[Alireza Mahdavi]

Hi Andrew;

Thank you. The answer was very helpful and worked with my code. My understanding is that I guess Stata does not understand numbers representing time in a format other than the number of milliseconds from a reference point.

Thanks
Alireza

[Petru Radulescu]

Hi everyone,

In this example how would you make the last 2 bars show the right way when crossing the 00:00 mark? In this example, they span the whole day even if the actual value is 23:30:00 - 01:59:59.
I have the same problem when plotting the start and end times of someone's sleep.

Thank you for the help!
Petru

[Andrew Musau]

As the clock resets in #2, you need a day variable to identify which day it is. While identifying the day is simple when the data span two days, dealing with data over multiple days may lead to issues if you attempt to infer days in the presence of multiple clock resets.

* Example generated by -dataex-. To install: ssc install dataex
clear
input float id double(starttime endtime)
 1 30600000 57600000
 2 43200000 48600000
 3 43200000 46800000
 4 48600000 54900000
 5 48600000 55799000
 6 50400000 54000000
 7 50400000 54000000
 8 52200000 55799000
 9 53100000 54000000
10 54000000 61199000
11 54000000 57600000
12 55800000 58499000
13 57600000 60300000
14 59400000 63000000
15 61200000 64800000
16 61200000 64800000
17 61200000 64800000
18 63000000 66599000
19 63000000 70200000
20 64800000 72900000
21 66600000 70200000
22 68400000 75600000
23 70200000 75600000
24 72000000 78300000
25 73800000 78000000
26 75600000 81000000
27 75600000 77399000
28 79200000 82799000
29 82800000  7199000
30 84600000  7199000
end
format %tcHH:MM:SS starttime
format %tcHH:MM:SS endtime

gen double new_endtime= cond(end<start, end+start, end)

tw (rcap starttime new_endtime id, xlab("") ylab(28800000(3600000)92000000, ///
format(%tcHH:MM:SS) angle(0))  xtitle("") ytitle(Start/End))

[Petru Radulescu]

Hi Andrew,

Thank you so much for the reply. This method worked very well, and with some adaption to the code, in the end I got this:

gen double new_starttime= cond(endtime<starttime, starttime - 3*endtime, starttime)

I pushed back the start time, which works better for my sleep time from 23:45:00 to 08:00:00.
The other way would show the bar for this somehow above all the other bars that span only one day.

Thank you for the help,
Petru