Sum row variables with Macro

[Francois Durant]

Hi,

I have the following dataset:

permno    date    ln_ret
10000    1986m1    
10000    1986m2    -.2972516
10000    1986m3    .3114361
10000    1986m4    -.1037968
10000    1986m5    -.2518726
10000    1986m6    -.0050378
10000    1986m7    -.0842603
10000    1986m8    -.9555115
10000    1986m9    -.0588405
10000    1986m10    -.2776318
10000    1986m11    .0582689
10000    1986m12    -.4737844
10001    1986m1    
10001    1986m2    .0202027
10001    1986m3    .0248877
10001    1986m4    .0098523
10001    1986m5    -.0098523
10001    1986m6    -.0131555
10001    1986m7    -.0102565
10001    1986m8    .0696799
10001    1986m9    -.0030817
10001    1986m10    .0384663
10001    1986m11    .0550598
10001    1986m12    .0148886

I would like to add the variables ret_L1+ret_L2+...+ret_L`J'. I defined the local macro `J' in the code. How can I modify the last row? Here is what I am running

local J 6

tsset permno date, monthly
sort permno date

forvalues i = 1(1)`J' {
gen ret_L`i' = L`i'.ln_ret
}

local v ret_L1-ret_L`J'

gen cum_ret = exp(sum(`v'))-1

[Nick Cox]

This won't work any way.

The function sum() gives you a cumulative sum across observations. You want a row sum across variables. Check out tsegen (SSC)

 
clear 
input permno  str7 s_date    ln_ret
10000    1986m1        .
10000    1986m2    -.2972516
10000    1986m3    .3114361
10000    1986m4    -.1037968
10000    1986m5    -.2518726
10000    1986m6    -.0050378
10000    1986m7    -.0842603
10000    1986m8    -.9555115
10000    1986m9    -.0588405
10000    1986m10    -.2776318
10000    1986m11    .0582689
10000    1986m12    -.4737844
10001    1986m1        . 
10001    1986m2    .0202027
10001    1986m3    .0248877
10001    1986m4    .0098523
10001    1986m5    -.0098523
10001    1986m6    -.0131555
10001    1986m7    -.0102565
10001    1986m8    .0696799
10001    1986m9    -.0030817
10001    1986m10    .0384663
10001    1986m11    .0550598
10001    1986m12    .0148886
end 
gen date = monthly(s_date, "YM") 
tsset permno date 

tsegen wanted = rowtotal(L(1/6).ln_ret, 6)

[Francois Durant]

Thanks for the response Nick. I would like to program in a flexible manner by using the macro `J' as described above. Would you be able to help incorporating that in your code?

I have tried the following:

local J 6
tsegen wanted = rowtotal(L(1/6).ln_ret, 6)

Thanks again!

[Nick Cox]

  
local J 6
tsegen wanted = rowtotal(L(1/`J').ln_ret, `J')

[Francois Durant]

Nick,

I was wondering how results can be slighly different using the two approach:

Code:

local J 3
gen cum_ret_test = exp(ln_ret + ret_L1 + ret_L2 + ret_L3)-1
tsegen cum_ret0 = rowtotal(L(1/`J').ln_ret, `J')
gen cum_ret1 = exp(ln_ret+cum_ret0)-1
drop cum_ret0

outcome:

cum_ret_test    cum_ret1
-.0480769    -.048077
-.3591549    -.3591549
-.7265625    -.7265625
-.6683418    -.6683417

[Nick Cox]

Floating point arithmetic...

Spell out that you want doubles if high order agreement is important to you.