Run simple regression over a subsample by "in"

[John Baxter]

Hello everyone, I am new to Stata and woud like your help. I have a sample of size 249, and I would like to run a regression over the first 50 observations. The only way I know is to create a sequence variable and use "if", and it works (the blue line). I was wondering if I can use "in" instead? (the red line)
When I run the following code, I got an error message "'50-start_T+1' invalid observation number"


clear all

import excel "data/xyz.xlsx", firstrow case(lower) clear

egen t1 = seq()

tsset t1
sca n1 = 65
sca start_T = 50

gen b_cons = .
gen b_x1 = .
gen b_x2 = .
gen y_f = .

forvalues i = `=start_T'/`=_N-n1' {
//reg y x1 x2 if t1<= `i' & t1> `i'-start_T // it works
reg y x1 x2 in `i'-start_T+1/`i' // it does not work; error
predict temp in `i+n1'
replace y_f = temp in `i+n1'
drop temp
}


I would like to know how to do it because "if" does not extend to more advanced models, but I hope "in" can.

[Carlo Lazzaro]

John:
I think you can safely use -in- as you surmise:

. sysuse auto.dta
(1978 Automobile Data)

. regress price mpg in 1/24

      Source |       SS           df       MS      Number of obs   =        24
-------------+----------------------------------   F(1, 22)        =      6.90
       Model |    64504736         1    64504736   Prob > F        =    0.0154
    Residual |   205717267        22  9350784.84   R-squared       =    0.2387
-------------+----------------------------------   Adj R-squared   =    0.2041
       Total |   270222003        23  11748782.7   Root MSE        =    3057.9

------------------------------------------------------------------------------
       price |      Coef.   Std. Err.      t    P>|t|     [95% Conf. Interval]
-------------+----------------------------------------------------------------
         mpg |  -364.3311   138.7153    -2.63   0.015    -652.0092   -76.65311
       _cons |   13521.27    2871.86     4.71   0.000     7565.402    19477.15
------------------------------------------------------------------------------

.

If you want to collect coefficients and the like, take a look at -statsby-.

[George Ford]

asreg makes quick work of these sorts of problems, storing the coefficients/etc for each round.

search asreg

[Nick Cox]

The problem with #1 is the problem with your previous thread: you want scalar calculations on the fly, which is possible but with different syntax. I can't test your code -- your spreadsheet isn't visible to me -- but I venture this rewriting, which is just one way to do it;

clear all

import excel "data/xyz.xlsx", firstrow case(lower) clear

gen t1 = _n
// egen t1 = seq()

tsset t1
sca n1 = 65
sca start_T = 50

gen b_cons = .
gen b_x1 = .
gen b_x2 = .
gen y_f = .

forvalues i = `=start_T'/`=_N-n1' {

    local j = `i' - start_T 1
    reg y x1 x2 in `j'/`i'

    local k = `i' + n1
    predict temp in `k'
    replace y_f = temp in `k'
    drop temp

}

[John Baxter]

Thank you everyone for help!! Nick, would you mind testing the code with the xlsx file I just attached and provide a modification on my previous code? Thank you!!

[Carlo Lazzaro]

John:
nobody on this list will ever open spreadsheets coming from unknown sources due to the risk of active contents.

[Nick Cox]

To paraphrase Carlo Lazzaro slightly, we ask that you don't post spreadsheets. See https://www.statalist.org/forums/help and especially #12. All posters are asked to read this before posting.

I should apologise in this sense: it may have appeared that I was asking to see the spreadsheet, but I wasn't. My comment was intended as a little wry, along the lines of "You're using a spreadsheet on your machine which naturally I can't see; hence I can't be confident that this is exactly right."

In any case, I have already offered a modification of your previous code!