using labmask in tsline

[charlie wong]

hello everyone

Somehow I could not get the desired results when using labmask in the following situation. I want to produce a graph with a formatted date axis (code 1 below). In my data my time variable is not directly the date variable but group of dates (to take out the gaps between some dates). Just wondering if there is a way to show a formatted date axis with the grouped date variable. I tried with code 2 below with no success. Grateful for any help.

*CODE 1
clear 
set obs 10
gen date=20000+_n*2 //dates with gaps

gen x=_n

tsset date
tsline x , tlabel(, format(%dm-CY))

*************
*CODE 2
clear 
set obs 10
gen date=20000+_n*2 //dates with gaps
egen day=group(date)

gen x=_n

tsset day
g sdate=string(date,"%dm-CY")
labmask day, values(sdate)

tsline x

[Nick Cox]

labmask is from the Stata Journal, as you are asked to explain.

. search labmask, sj

Search of official help files, FAQs, Examples, SJs, and STBs

SJ-8-2  gr0034  . . . . . . . . . .  Speaking Stata: Between tables and graphs
        (help labmask, seqvar if installed) . . . . . . . . . . . .  N. J. Cox
        Q2/08   SJ 8(2):269--289
        outlines techniques for producing table-like graphs

I think your problem is with tsline which will respect an assigned display format, but doesn't by default look for value labels. Try further

tsline x, xla(, valuelabel)
tsline x, xla(1(2)9, valuelabel)

[charlie wong]

Thank you for the help, Nick!

[Nicu Sprincean]

Nick Cox
On my x-axis, I have the dates in format like DaymonthYear. When I use -tsline-, I want to plot only year. I've created a variable 'year' and applied the following code:
labmask(date), values(year)
tsline var, xlabel(,valuelabel).

But the results are quite surprising: the first date is in format like 03jan2000, the other is OK (year 2002), followed by other variable like 03jan2004 and again by the year (2006), and so on. Thank you.

[Nick Cox]

Friedrich Mises You don't give a reproducible data example that I or anyone else can play with. But this doesn't seem puzzling to me.

With tsline and no values specified you are relying on its defaultchoice of axis label positions. You will get value labels shown if and only if value labels are defined for the values chosen by tsline; if a value has no value label defined the display format you chose will come into effect.

If you run this code, you will see the principle exemplified.

clear
set obs 2 
gen date = cond(_n == 1, mdy(1,1,1998), mdy(1,1,2000))
gen year = year(date) 
format date %td
labmask date, values(year) 
gen y = 42
tsset date 
tsline y , xla(, valuelabels)

A more direct solution for what you want is likely to be

tsline var, xlabel(, format(%tdCCYY)) 

but odd results are quite possible unless your axis labels are an exact number of years apart.

[Nicu Sprincean]

Nick Cox, thank you very much, the last code worked just fine.