Multiple fixed effects using areg, absorb(fe)

[Syed Raza]

Hello everyone,

I am having three different regions for my regression analysis. Every region consists of distinct countries. I am running regression region wise and everything is looped over. Now, for example, I have 20 countries (and i want to use country wise fixed effects) thus have generated 20 dummies for all 20 countries which means 20 columns consist of 0 and 1. Here is the command i am using for regression

areg `v' evt_time evt_l* `controls' [pweight=rwtresp], absorb(dum*)  cluster(hhid)

for v local is defined where i have all the variables of interest. Since i have dum1 dum2.....dum20. That's how i am trying to use dummies with absorb. It works fine if i use only one 1 dummy like dum15 or dum2. But it does not work altogether in the current format and gives me this error

absorb(): too many variables specified

Can you please suggest how to fix this problem.

[Carlo Lazzaro]

Syed:
what you're experiencing is related to -areg- mechanism.
-absorb()- should be fed with one categorical variable only, as you can see from the following toy-example:

. sysuse auto
(1978 Automobile Data)

. areg price weight length, absorb(rep78)

Linear regression, absorbing indicators         Number of obs     =         69
Absorbed variable: rep78                        No. of categories =          5
                                                F(   2,     62)   =      22.98
                                                Prob > F          =     0.0000
                                                R-squared         =     0.4341
                                                Adj R-squared     =     0.3793
                                                Root MSE          =  2294.5106

------------------------------------------------------------------------------
       price |      Coef.   Std. Err.      t    P>|t|     [95% Conf. Interval]
-------------+----------------------------------------------------------------
      weight |   5.478309   1.158582     4.73   0.000     3.162337    7.794281
      length |  -109.5065   39.26104    -2.79   0.007    -187.9882   -31.02482
       _cons |   10154.62   4270.525     2.38   0.021      1617.96    18691.27
------------------------------------------------------------------------------
F test of absorbed indicators: F(4, 62) = 2.079               Prob > F = 0.094

. tab rep78, g(dummy)

     Repair |
Record 1978 |      Freq.     Percent        Cum.
------------+-----------------------------------
          1 |          2        2.90        2.90
          2 |          8       11.59       14.49
          3 |         30       43.48       57.97
          4 |         18       26.09       84.06
          5 |         11       15.94      100.00
------------+-----------------------------------
      Total |         69      100.00

. areg price weight length, absorb(dummy*)
absorb():  too many variables specified
r(103);

. g categorical_rep78=1 if dummy1==1
(72 missing values generated)

. replace categorical_rep78=2 if dummy2==1
(8 real changes made)

. replace categorical_rep78=3 if dummy3==1
(30 real changes made)

. replace categorical_rep78=4 if dummy4==1
(18 real changes made)

. replace categorical_rep78=5 if dummy5==1
(11 real changes made)

. areg price weight length, absorb( categorical_rep78 )

Linear regression, absorbing indicators         Number of obs     =         69
Absorbed variable: categorical_rep78            No. of categories =          5
                                                F(   2,     62)   =      22.98
                                                Prob > F          =     0.0000
                                                R-squared         =     0.4341
                                                Adj R-squared     =     0.3793
                                                Root MSE          =  2294.5106

------------------------------------------------------------------------------
       price |      Coef.   Std. Err.      t    P>|t|     [95% Conf. Interval]
-------------+----------------------------------------------------------------
      weight |   5.478309   1.158582     4.73   0.000     3.162337    7.794281
      length |  -109.5065   39.26104    -2.79   0.007    -187.9882   -31.02482
       _cons |   10154.62   4270.525     2.38   0.021      1617.96    18691.27
------------------------------------------------------------------------------
F test of absorbed indicators: F(4, 62) = 2.079               Prob > F = 0.094

Hence, you have to create am unique 20-level categorical variable and plug it in as the argument of -absorb()-.

[Syed Raza]

You are right I just figure that out. In addition to this, I have already created 20 unique variables I used local for dummies where i

local fes "dum1 dum2 dum3 dum4 dum5 dum6 dum7 dum8 dum9 dum10 dum11 dum12 dum13 dum14 dum15 dum16 dum17 dum18 dum19 dum20"
local outcomes "oop_spend age_hosp working ritearnsemp sitearnsemp hgovt hittot log_hittot"


foreach fe in `fes' {

foreach v of varlist `outcomes' { 

    local controls ""
    if "`fe'"=="hhidpn" {
        xi i.wave
        drop _Iwave_11
        local controls "_I*"
        if regexm("`v'","_c")==1 {
            drop _Iwave_2 _Iwave_3 
        }
        
        drop _Iwave_2 _Iwave_3 
    }



areg `v' evt_time evt_l* `controls' [pweight=rwtresp], absorb(`fes')  cluster(hhidpn)

}
}

but still, the error remains the same. I don't know why?

[Syed Raza]

I want to use argument because i want to add one more fixed effect (cohortXwave) in my regression in addition to country fixed effects. In that case, it will not work even if i recode all the dummies into one categorical variable.

[Carlo Lazzaro]

Syed:
- I cannot say what went wrong with your code;
. as far as your last question is concerned, you can take a look at the user-written programme -reghdfe- by Sergio Correia.

[Syed Raza]

solved thanks

[Carlo Lazzaro]

Syed:
happy with reading that you've found a solution.
You may want to share it with the list, as other could benefit from it.