[Giorgio Di Stefano]

Thank you, Andrew , for one more time!

Sorry if I am insisting too much, but I am not getting the last part
So, given that in my case

nlcom -_b[indicator]/(2*_b[indicator#indicator])

I wrote the code as follows in order to get 1-a .

foreach var of varlist decade?{
xtset id ts
xtregar D.gdp cpi u dummy1 dummy2 c.indicator##c.indicator if `var', fe
// get the x where the minimum occurs
local x = -_b[indicator]/(2*_b[indicator1#indicator])
// get the y for that minimim
qui margins, at(indicator = `x' cpi=0 u=2 dummy1=0 dummy2=0 indicator=-4)
// the predictions are stored in the matrix r(b)
// el(r(b),1,1) extracts the cell 1,1 from the matrix r(b)
local y = el(r(b),1,1)
gen indmax2_`var'= `x'
gen gdpmax2_`var'=`y'
gen a_`var' = -_b[indicator]/(2*_b[indicator#indicator])
gen dep_`var' = 1-a
// prepare for our plot
nlcom -_b[indicator]/(2*_b[indicator#indicator])

qui margins, at(indicator=(-16(4)29) cpi=0 u=2 dummy1=0 dummy2=0 indicator=-4)
// with scatteri we can add a point and a label to our graph
marginsplot, noci plotopts(msymbol(i)) legend(off) ///
addplot(scatteri `y' `x' (12) "(`xlab'; `ylab')") ///
ylab(,format(%9.3fc) angle(0)) ytitle("predicted gdp ") title("Predicted Marings pair")
}

It does work for the entire sample, but again It fails when I add the if option, in the regression .I get no observation. What did I do wrong?
It is similar to what I had referred in #14

[Andrew Musau]

Exit the loop and investigate why the regression is failing, e.g., by running

xtset id ts
xtregar D.gdp cpi u dummy1 dummy2 c.indicator##c.indicator if decade1, fe

You are using a panel data estimator and you have a differenced variable. Differencing results in the loss of one cross-section assuming a balanced panel as \(\Delta \text{ GDP}_t\;= \text{ GDP}_t\;-\;\text{ GDP}_{t-1}\), or the current level of GDP minus its one-period lag. For a fixed effects model, you need a minimum of 2 observations, so with the differencing, this becomes 3. With an unbalanced panel, perhaps you are left with no observations in some decades after differencing. Or it could be something else.

[Giorgio Di Stefano]

Exit the loop and investigate why the regression is failing, e.g., by running

xtset id ts
xtregar D.gdp cpi u dummy1 dummy2 c.indicator##c.indicator if decade1, fe

You are using a panel data estimator and you have a differenced variable. Differencing results in the loss of one cross-section assuming a balanced panel as \(\Delta \text{ GDP}_t\;= \text{ GDP}_t\;-\;\text{ GDP}_{t-1}\), or the current level of GDP minus its one-period lag. For a fixed effects model, you need a minimum of 2 observations, so with the differencing, this becomes 3. With an unbalanced panel, perhaps you are left with no observations in some decades after differencing. Or it could be something else.

Now I got it! I am so grateful to you for your help and time!
May I add a last query before endings this thread.
How can I include in the loop a margins save plot for each graph produced, and secondly produce pubbication tables by entire period, subperiods, subgroups etc, including the
(1-a) defined as 1-nlcom in #16
Thank you so much

[Andrew Musau]

For saving and exporting graphs, see

help gr export

You need to name the graphs differently so that some are not overwritten. For storing estimates, see

help estimates store

You can use the same prefixes to make things easier, e.g.,

foreach var of varlist decade?{
xtset id ts
...
xtregar D.gdp cpi u dummy1 dummy2 c.indicator##c.indicator if `var', fe
est store est_`var'
...
marginsplot, noci plotopts(msymbol(i)) legend(off) ///
addplot(scatteri `y' `x' (12) "(`xlab'; `ylab')") ///
ylab(,format(%9.3fc) angle(0)) ytitle("predicted gdp ") title("Predicted Marings pair: `var'")
gr export gr_`var', as(png) replace
}

Then with a command such as estout from SSC

esttab est*

[Giorgio Di Stefano]

For saving and exporting graphs, see

help gr export

You need to name the graphs differently so that some are not overwritten. For storing estimates, see

help estimates store

You can use the same prefixes to make things easier, e.g.,

foreach var of varlist decade?{
xtset id ts
...
xtregar D.gdp cpi u dummy1 dummy2 c.indicator##c.indicator if `var', fe
est store est_`var'
...
marginsplot, noci plotopts(msymbol(i)) legend(off) ///
addplot(scatteri `y' `x' (12) "(`xlab'; `ylab')") ///
ylab(,format(%9.3fc) angle(0)) ytitle("predicted gdp ") title("Predicted Marings pair: `var'")
gr export gr_`var', as(png) replace
}

Then with a command such as estout from SSC

esttab est*

Thank you very much, Andrew! I had a such thing in mind, ;but was not sure about the loop. Really, you helped me a lot. Two last thing easy to answers, I promise.

1. I need to derive (collapse) the weighted average of their each entire period and by groups, and then for each decade and for each group. Including the (1-a) values defined as 1-nlcom in #16

and

2. Will the loop still work if I substitute the indicators with a local say called ind

local Ind indicator1 indicator2 etc


as I have 12 of them, and some have different names? Same question for the margins

[Andrew Musau]

I do not follow #1. As I said, I do not understand what you are doing. You can post in a new thread and see if there is anyone that can follow. For #2, yes, you can store variables in locals and call them in a loop. Here is one way to do it:

sysuse auto, clear
set seed 03222022
forval i=1/4{
    gen indicator`i'= runiformint(0,1)
    gen dummy`i'= runiformint(0,1)
}
local indicators indicator1 indicator2 indicator3 indicator4
local dummies dummy1 dummy2 dummy3 dummy4
foreach var in `indicators' `dummies' {
    display "Regression for `var'"
    regress mpg weight if `var'
}

Res.:

Regression for indicator1

      Source |       SS           df       MS      Number of obs   =        37
-------------+----------------------------------   F(1, 35)        =     78.03
       Model |  812.742905         1  812.742905   Prob > F        =    0.0000
    Residual |  364.554393        35  10.4158398   R-squared       =    0.6903
-------------+----------------------------------   Adj R-squared   =    0.6815
       Total |   1177.2973        36  32.7027027   Root MSE        =    3.2274

------------------------------------------------------------------------------
         mpg |      Coef.   Std. Err.      t    P>|t|     [95% Conf. Interval]
-------------+----------------------------------------------------------------
      weight |  -.0061288   .0006938    -8.83   0.000    -.0075373   -.0047203
       _cons |   39.84612   2.118409    18.81   0.000     35.54552    44.14672
------------------------------------------------------------------------------
Regression for indicator2

      Source |       SS           df       MS      Number of obs   =        29
-------------+----------------------------------   F(1, 27)        =     55.79
       Model |  641.898254         1  641.898254   Prob > F        =    0.0000
    Residual |   310.65347        27  11.5056841   R-squared       =    0.6739
-------------+----------------------------------   Adj R-squared   =    0.6618
       Total |  952.551724        28  34.0197044   Root MSE        =     3.392

------------------------------------------------------------------------------
         mpg |      Coef.   Std. Err.      t    P>|t|     [95% Conf. Interval]
-------------+----------------------------------------------------------------
      weight |  -.0056985   .0007629    -7.47   0.000    -.0072639   -.0041331
       _cons |   39.68576   2.365498    16.78   0.000     34.83216    44.53936
------------------------------------------------------------------------------
Regression for indicator3

      Source |       SS           df       MS      Number of obs   =        40
-------------+----------------------------------   F(1, 38)        =     58.11
       Model |  858.289674         1  858.289674   Prob > F        =    0.0000
    Residual |  561.310326        38  14.7713244   R-squared       =    0.6046
-------------+----------------------------------   Adj R-squared   =    0.5942
       Total |      1419.6        39        36.4   Root MSE        =    3.8433

------------------------------------------------------------------------------
         mpg |      Coef.   Std. Err.      t    P>|t|     [95% Conf. Interval]
-------------+----------------------------------------------------------------
      weight |  -.0063738   .0008362    -7.62   0.000    -.0080666   -.0046811
       _cons |   40.89364   2.603014    15.71   0.000     35.62411    46.16316
------------------------------------------------------------------------------
Regression for indicator4

      Source |       SS           df       MS      Number of obs   =        40
-------------+----------------------------------   F(1, 38)        =     69.10
       Model |   729.01922         1   729.01922   Prob > F        =    0.0000
    Residual |   400.88078        38  10.5494942   R-squared       =    0.6452
-------------+----------------------------------   Adj R-squared   =    0.6359
       Total |      1129.9        39  28.9717949   Root MSE        =     3.248

------------------------------------------------------------------------------
         mpg |      Coef.   Std. Err.      t    P>|t|     [95% Conf. Interval]
-------------+----------------------------------------------------------------
      weight |  -.0058179   .0006999    -8.31   0.000    -.0072347   -.0044011
       _cons |   38.79975   2.313221    16.77   0.000     34.11688    43.48262
------------------------------------------------------------------------------
Regression for dummy1

      Source |       SS           df       MS      Number of obs   =        40
-------------+----------------------------------   F(1, 38)        =     68.35
       Model |  929.289738         1  929.289738   Prob > F        =    0.0000
    Residual |  516.685262        38  13.5969806   R-squared       =    0.6427
-------------+----------------------------------   Adj R-squared   =    0.6333
       Total |    1445.975        39  37.0762821   Root MSE        =    3.6874

------------------------------------------------------------------------------
         mpg |      Coef.   Std. Err.      t    P>|t|     [95% Conf. Interval]
-------------+----------------------------------------------------------------
      weight |  -.0062697   .0007584    -8.27   0.000     -.007805   -.0047344
       _cons |   40.58582   2.354745    17.24   0.000     35.81888    45.35275
------------------------------------------------------------------------------
Regression for dummy2

      Source |       SS           df       MS      Number of obs   =        32
-------------+----------------------------------   F(1, 30)        =     39.30
       Model |  583.545776         1  583.545776   Prob > F        =    0.0000
    Residual |  445.422974        30  14.8474325   R-squared       =    0.5671
-------------+----------------------------------   Adj R-squared   =    0.5527
       Total |  1028.96875        31  33.1925403   Root MSE        =    3.8532

------------------------------------------------------------------------------
         mpg |      Coef.   Std. Err.      t    P>|t|     [95% Conf. Interval]
-------------+----------------------------------------------------------------
      weight |  -.0054957   .0008766    -6.27   0.000     -.007286   -.0037054
       _cons |   37.93845   2.791229    13.59   0.000       32.238     43.6389
------------------------------------------------------------------------------
Regression for dummy3

      Source |       SS           df       MS      Number of obs   =        39
-------------+----------------------------------   F(1, 37)        =     77.01
       Model |  466.687495         1  466.687495   Prob > F        =    0.0000
    Residual |  224.235582        37  6.06042113   R-squared       =    0.6755
-------------+----------------------------------   Adj R-squared   =    0.6667
       Total |  690.923077        38  18.1821862   Root MSE        =    2.4618

------------------------------------------------------------------------------
         mpg |      Coef.   Std. Err.      t    P>|t|     [95% Conf. Interval]
-------------+----------------------------------------------------------------
      weight |  -.0045875   .0005228    -8.78   0.000    -.0056467   -.0035282
       _cons |   34.61545   1.685958    20.53   0.000     31.19937    38.03152
------------------------------------------------------------------------------
Regression for dummy4

      Source |       SS           df       MS      Number of obs   =        34
-------------+----------------------------------   F(1, 32)        =     55.05
       Model |  843.788171         1  843.788171   Prob > F        =    0.0000
    Residual |  490.447123        32  15.3264726   R-squared       =    0.6324
-------------+----------------------------------   Adj R-squared   =    0.6209
       Total |  1334.23529        33  40.4313725   Root MSE        =    3.9149

------------------------------------------------------------------------------
         mpg |      Coef.   Std. Err.      t    P>|t|     [95% Conf. Interval]
-------------+----------------------------------------------------------------
      weight |  -.0065315   .0008803    -7.42   0.000    -.0083246   -.0047385
       _cons |   41.61033   2.650133    15.70   0.000     36.21219    47.00848
------------------------------------------------------------------------------

.

[Giorgio Di Stefano]

I do not follow #1. As I said, I do not understand what you are doing. You can post in a new thread and see if there is anyone that can follow. For #2, yes, you can store variables in locals and call them in a loop. Here is one way to do it:

sysuse auto, clear
set seed 03222022
forval i=1/4{
gen indicator`i'= runiformint(0,1)
gen dummy`i'= runiformint(0,1)
}
local indicators indicator1 indicator2 indicator3 indicator4
local dummies dummy1 dummy2 dummy3 dummy4
foreach var in `indicators' `dummies' {
display "Regression for `var'"
regress mpg weight if `var'
}

Res.:

Regression for indicator1

Source | SS df MS Number of obs = 37
-------------+---------------------------------- F(1, 35) = 78.03
Model | 812.742905 1 812.742905 Prob > F = 0.0000
Residual | 364.554393 35 10.4158398 R-squared = 0.6903
-------------+---------------------------------- Adj R-squared = 0.6815
Total | 1177.2973 36 32.7027027 Root MSE = 3.2274

------------------------------------------------------------------------------
mpg | Coef. Std. Err. t P>|t| [95% Conf. Interval]
-------------+----------------------------------------------------------------
weight | -.0061288 .0006938 -8.83 0.000 -.0075373 -.0047203
_cons | 39.84612 2.118409 18.81 0.000 35.54552 44.14672
------------------------------------------------------------------------------
Regression for indicator2

Source | SS df MS Number of obs = 29
-------------+---------------------------------- F(1, 27) = 55.79
Model | 641.898254 1 641.898254 Prob > F = 0.0000
Residual | 310.65347 27 11.5056841 R-squared = 0.6739
-------------+---------------------------------- Adj R-squared = 0.6618
Total | 952.551724 28 34.0197044 Root MSE = 3.392

------------------------------------------------------------------------------
mpg | Coef. Std. Err. t P>|t| [95% Conf. Interval]
-------------+----------------------------------------------------------------
weight | -.0056985 .0007629 -7.47 0.000 -.0072639 -.0041331
_cons | 39.68576 2.365498 16.78 0.000 34.83216 44.53936
------------------------------------------------------------------------------
Regression for indicator3

Source | SS df MS Number of obs = 40
-------------+---------------------------------- F(1, 38) = 58.11
Model | 858.289674 1 858.289674 Prob > F = 0.0000
Residual | 561.310326 38 14.7713244 R-squared = 0.6046
-------------+---------------------------------- Adj R-squared = 0.5942
Total | 1419.6 39 36.4 Root MSE = 3.8433

------------------------------------------------------------------------------
mpg | Coef. Std. Err. t P>|t| [95% Conf. Interval]
-------------+----------------------------------------------------------------
weight | -.0063738 .0008362 -7.62 0.000 -.0080666 -.0046811
_cons | 40.89364 2.603014 15.71 0.000 35.62411 46.16316
------------------------------------------------------------------------------
Regression for indicator4

Source | SS df MS Number of obs = 40
-------------+---------------------------------- F(1, 38) = 69.10
Model | 729.01922 1 729.01922 Prob > F = 0.0000
Residual | 400.88078 38 10.5494942 R-squared = 0.6452
-------------+---------------------------------- Adj R-squared = 0.6359
Total | 1129.9 39 28.9717949 Root MSE = 3.248

------------------------------------------------------------------------------
mpg | Coef. Std. Err. t P>|t| [95% Conf. Interval]
-------------+----------------------------------------------------------------
weight | -.0058179 .0006999 -8.31 0.000 -.0072347 -.0044011
_cons | 38.79975 2.313221 16.77 0.000 34.11688 43.48262
------------------------------------------------------------------------------
Regression for dummy1

Source | SS df MS Number of obs = 40
-------------+---------------------------------- F(1, 38) = 68.35
Model | 929.289738 1 929.289738 Prob > F = 0.0000
Residual | 516.685262 38 13.5969806 R-squared = 0.6427
-------------+---------------------------------- Adj R-squared = 0.6333
Total | 1445.975 39 37.0762821 Root MSE = 3.6874

------------------------------------------------------------------------------
mpg | Coef. Std. Err. t P>|t| [95% Conf. Interval]
-------------+----------------------------------------------------------------
weight | -.0062697 .0007584 -8.27 0.000 -.007805 -.0047344
_cons | 40.58582 2.354745 17.24 0.000 35.81888 45.35275
------------------------------------------------------------------------------
Regression for dummy2

Source | SS df MS Number of obs = 32
-------------+---------------------------------- F(1, 30) = 39.30
Model | 583.545776 1 583.545776 Prob > F = 0.0000
Residual | 445.422974 30 14.8474325 R-squared = 0.5671
-------------+---------------------------------- Adj R-squared = 0.5527
Total | 1028.96875 31 33.1925403 Root MSE = 3.8532

------------------------------------------------------------------------------
mpg | Coef. Std. Err. t P>|t| [95% Conf. Interval]
-------------+----------------------------------------------------------------
weight | -.0054957 .0008766 -6.27 0.000 -.007286 -.0037054
_cons | 37.93845 2.791229 13.59 0.000 32.238 43.6389
------------------------------------------------------------------------------
Regression for dummy3

Source | SS df MS Number of obs = 39
-------------+---------------------------------- F(1, 37) = 77.01
Model | 466.687495 1 466.687495 Prob > F = 0.0000
Residual | 224.235582 37 6.06042113 R-squared = 0.6755
-------------+---------------------------------- Adj R-squared = 0.6667
Total | 690.923077 38 18.1821862 Root MSE = 2.4618

------------------------------------------------------------------------------
mpg | Coef. Std. Err. t P>|t| [95% Conf. Interval]
-------------+----------------------------------------------------------------
weight | -.0045875 .0005228 -8.78 0.000 -.0056467 -.0035282
_cons | 34.61545 1.685958 20.53 0.000 31.19937 38.03152
------------------------------------------------------------------------------
Regression for dummy4

Source | SS df MS Number of obs = 34
-------------+---------------------------------- F(1, 32) = 55.05
Model | 843.788171 1 843.788171 Prob > F = 0.0000
Residual | 490.447123 32 15.3264726 R-squared = 0.6324
-------------+---------------------------------- Adj R-squared = 0.6209
Total | 1334.23529 33 40.4313725 Root MSE = 3.9149

------------------------------------------------------------------------------
mpg | Coef. Std. Err. t P>|t| [95% Conf. Interval]
-------------+----------------------------------------------------------------
weight | -.0065315 .0008803 -7.42 0.000 -.0083246 -.0047385
_cons | 41.61033 2.650133 15.70 0.000 36.21219 47.00848
------------------------------------------------------------------------------

.

Thank you again for everything, Andrew! I truly appreciate your help and time!
Thanks a ton!

Giorgio!