Removing multiple substrings from macros

[Jared Greathouse]

I wanna remove multiple subtrings from a macro.

qui ssc inst rcm, replace


import delim "https://raw.githubusercontent.com/jehangiramjad/tslib/master/tests/testdata/prop99.csv", clear

loc int_time = 1989

drop if year >= 2001


keep if submeasureiddisplayorder==2

rename (locationdesc data_value) (state cigsale)

keep state year cig

egen id = group(state)

order id, b(year)

drop if inlist(id,3,22,38,10,12,2,12,23,31,33,48)

xtset id year, y
cls

rcm cig, ///
    trunit(5) ///
    trperiod(`int_time') ///
    method(forward) ///
    cr(mbic) frame(cig_wide)


local sel = subinstr("`e(predvar_sel)'","cigsale", """,",.)

di "`sel'"

The final macro looks like

·43 ·28 ·47 ·1 ·24 ·44 ·4 ·40 ·41 ·6 ·27 ·45 ·25 ·29 ·14 ·30

and I desire

,43,28,47,1,24,44,4

and so on. How might I do this?

Methodologically, I like the forward selection method, but in my personal opinion it WILDLY overfits the time-series in the pre-intervention period thus resulting in a noisy counterfactual. The author tells me that the method doesn't play well with non-stationary time series data, so I wanna see if I can fix it.


Nevermind, I got it

local sel = subinstr("`e(predvar_sel)'","cigsale·",","",",.)

[Jared Greathouse]

Actually, no I don't.. Consider this simple example

cls

loc string gdpcap·4 gdpcap·445434


local sel = subinstr("`string'","gdpcap·" , ",",.)


di `sel'

I want 4,445434

but instead, there are spaces between them.

[Andrew Musau]

loc string gdpcap·4 gdpcap·445434·43 ·28 ·47 ·1 ·24 ·44 ·4 ·40 ·41 ·6 ·27 ·45 ·25 ·29 ·14 ·30
local string= subinstr(ustrregexra("`string'", "[^\d]+", ","), ",","", 1)

Res.:

. di "`string'"
4,445434,43,28,47,1,24,44,4,40,41,6,27,45,25,29,14,30

[Jared Greathouse]

Yep this works. Thank you so much!