How to generate this formula

[Curtis Chou]

Hello, I need help from every expert, I want to generate this formula. How to use "rangestat" to make it or other ways.

(sorry, the word is a little tie, so I expand the fig)

* Example generated by -dataex-. To install: ssc install dataex
clear
input float t double sum_sentiment
 1  .0288127515759094
 2 -.0904761904761904
 3   .113189484126984
 4   .143840852130325
 5   .214285714285714
 6  .0275213675213675
 7  -.333333333333333
 8  .0197083794758213
 9  .0621107331821617
10  .0865143862871135
11  .0175470219435736
12   .115094302398989
13  .0286485545106234
14  .0687808172914556
15  .0211010709504685
16   .109090909090909
17   .119491470830756
18                -.2
19   .133333333333333
20                -.1
end

[Nick Cox]

Isn't this just exponential smoothing as implemented by tssmooth?

Still, this is an attempt using rangerun from SSC. You may wish to discard results based on incomplete windows.

* Example generated by -dataex-. To install: ssc install dataex
clear
input float t double sum_sentiment
 1  .0288127515759094
 2 -.0904761904761904
 3   .113189484126984
 4   .143840852130325
 5   .214285714285714
 6  .0275213675213675
 7  -.333333333333333
 8  .0197083794758213
 9  .0621107331821617
10  .0865143862871135
11  .0175470219435736
12   .115094302398989
13  .0286485545106234
14  .0687808172914556
15  .0211010709504685
16   .109090909090909
17   .119491470830756
18                -.2
19   .133333333333333
20                -.1
end

capture program drop myprog 

program myprog 
    gen double wtsum = sum_sentiment[_N]
    gen count = _N 
    local I = _N - 1 
    forval i = 1/`I' { 
        replace wtsum = wtsum + sum_sentiment[_N - `i'] * exp(`i'/7)
    } 
end 

rangerun myprog    , use(sum_sentiment) interval(t -6 0)

list

     +--------------------------------------+
     |  t   sum_sent~t        wtsum   count |
     |--------------------------------------|
  1. |  1    .02881275    .02881275       1 |
  2. |  2   -.09047619   -.05723881       2 |
  3. |  3    .11318948     .0471608       3 |
  4. |  4    .14384085     .1982439       4 |
  5. |  5    .21428571    .44297294       5 |
     |--------------------------------------|
  6. |  6    .02752137    .53851944       6 |
  7. |  7   -.33333333    .28788384       7 |
  8. |  8    .01970838    .27347992       7 |
  9. |  9    .06211073    .62352738       7 |
 10. | 10    .08651439    .49811283       7 |
     |--------------------------------------|
 11. | 11    .01754702    .20115257       7 |
 12. | 12     .1150943    -.2353521       7 |
 13. | 13    .02864855   -.31765622       7 |
 14. | 14    .06878082    .60843767       7 |
 15. | 15    .02110107    .66940054       7 |
     |--------------------------------------|
 16. | 16    .10909091    .71245346       7 |
 17. | 17    .11949147    .70618235       7 |
 18. | 18          -.2    .56692949       7 |
 19. | 19    .13333333     .4744646       7 |
 20. | 20          -.1    .36945091       7 |
     +--------------------------------------+

[Curtis Chou]

thank you professor

[Curtis Chou]

Hello professor
May I inquire some questions? Your code is wonderful.
However, I want to learn what's means.

First, "gen double wtsum = sum_sentiment[_N]" , it would run a zero vector, why?

Second, local I = _N - 1
forval i = 1/`I' {
replace wtsum = wtsum + sum_sentiment[_N - `i'] * exp(`i'/7)
}

end
it would run a 1.7322000e+101 of vector, why?

[Nick Cox]

Sorry I don't understand what you're asking in #4 at all.

[Curtis Chou]

I am sorry, I forgot the dataset is different
OK repeat one again


First, "gen double wtsum = sum_sentiment[_N]" , it would run a -.1 vector, why?

Second, local I = _N - 1
forval i = 1/`I' {
replace wtsum = wtsum + sum_sentiment[_N - `i'] * exp(`i'/7)
}

end
it would run a 3.485122131255 of vector, why?

[Nick Cox]

Sorry, I fail again to follow what you are asking.

[Curtis Chou]

Why the ""gen double wtsum = sum_sentiment[_N]" can generate a vector about -.1?

[Nick Cox]

If you use that command outside of rangerun it records the last value in the dataset.

When rangerun issues that command within a program it applies only to the window of observations being analysed at any one point.