R-squared and -didregress-

[Michael Wharff]

Good evening all,

Is there a way to have -didregress- show a value for r-squared- as you would see if you did a difference-in-difference regression using the old methods (just using -reg-)?

Thanks!

[Andrew Musau]

You can quietly run the regression using regress and obtain the statistic. This code is valid for a single variable in -group()-.

use https://www.stata-press.com/data/r17/hospdd.dta, clear
didregress (satis)(procedure), group(hospital) time(month)
qui regress `e(depvar)' `e(treatment)' i.`e(clustvar)' i.`e(timevar)' if e(sample), cluster(`e(clustvar)')
display e(r2)

Res.:

. didregress (satis)(procedure), group(hospital) time(month)

Number of groups and treatment time

Time variable: month
Control:       procedure = 0
Treatment:     procedure = 1
-----------------------------------
             |   Control  Treatment
-------------+---------------------
Group        |
    hospital |        28         18
-------------+---------------------
Time         |
     Minimum |         1          4
     Maximum |         1          4
-----------------------------------

Difference-in-differences regression                     Number of obs = 7,368
Data type: Repeated cross-sectional

                               (Std. err. adjusted for 46 clusters in hospital)
-------------------------------------------------------------------------------
              |               Robust
        satis | Coefficient  std. err.      t    P>|t|     [95% conf. interval]
--------------+----------------------------------------------------------------
ATET          |
    procedure |
(New vs Old)  |   .8479879   .0321121    26.41   0.000     .7833108     .912665
-------------------------------------------------------------------------------
Note: ATET estimate adjusted for group effects and time effects.

. 
. qui regress `e(depvar)' `e(treatment)' i.`e(clustvar)' i.`e(timevar)' if e(sample), cluster(`e(clustvar)')

. 
. display e(r2)
.53325215

.

[Michael Wharff]

Thanks for the help Andrew! I do have a slight issue when running this command though. This works when the cluster variable has a small amount of values. However, my group variable is made up of 52,317 clusters. As such, when I run the command, I get the r(950) insufficient memory error. I have 44.6 million observations and Stats uses 5.7GB of memory. I have allowed Stats to use all available memory which is ~64GB total. I have also allowed Stata to use the maximum of 120,000 variables for State MP. I'm not sure if I'll be able to get around this one...

Thanks,

[Andrew Musau]

Install reghdfe from SSC.

ssc install reghdfe, replace
net install ftools, from("https://github.com/sergiocorreia/ftools/raw/master/src/")

use https://www.stata-press.com/data/r17/hospdd.dta, clear
didregress (satis)(procedure), group(hospital) time(month)
reghdfe `e(depvar)' `e(treatment)', absorb(`e(clustvar)' `e(timevar)') cluster(`e(clustvar)')
di e(r2)

Res.:

. didregress (satis)(procedure), group(hospital) time(month)

Number of groups and treatment time

Time variable: month
Control:       procedure = 0
Treatment:     procedure = 1
-----------------------------------
             |   Control  Treatment
-------------+---------------------
Group        |
    hospital |        28         18
-------------+---------------------
Time         |
     Minimum |         1          4
     Maximum |         1          4
-----------------------------------

Difference-in-differences regression                     Number of obs = 7,368
Data type: Repeated cross-sectional

                               (Std. err. adjusted for 46 clusters in hospital)
-------------------------------------------------------------------------------
              |               Robust
        satis | Coefficient  std. err.      t    P>|t|     [95% conf. interval]
--------------+----------------------------------------------------------------
ATET          |
    procedure |
(New vs Old)  |   .8479879   .0321121    26.41   0.000     .7833108     .912665
-------------------------------------------------------------------------------
Note: ATET estimate adjusted for group effects and time effects.

. 
. reghdfe `e(depvar)' `e(treatment)', absorb(`e(clustvar)' `e(timevar)') cluster(`e(clustvar)')
(MWFE estimator converged in 2 iterations)

HDFE Linear regression                            Number of obs   =      7,368
Absorbing 2 HDFE groups                           F(   1,     45) =     701.63
Statistics robust to heteroskedasticity           Prob > F        =     0.0000
                                                  R-squared       =     0.5333
                                                  Adj R-squared   =     0.5299
                                                  Within R-sq.    =     0.0774
Number of clusters (hospital) =         46        Root MSE        =     0.7239

                              (Std. err. adjusted for 46 clusters in hospital)
------------------------------------------------------------------------------
             |               Robust
       satis | Coefficient  std. err.      t    P>|t|     [95% conf. interval]
-------------+----------------------------------------------------------------
   procedure |   .8479879   .0320138    26.49   0.000     .7835088    .9124669
       _cons |   3.442755   .0066565   517.20   0.000     3.429348    3.456162
------------------------------------------------------------------------------

Absorbed degrees of freedom:
-----------------------------------------------------+
 Absorbed FE | Categories  - Redundant  = Num. Coefs |
-------------+---------------------------------------|
    hospital |        46          46           0    *|
       month |         7           0           7     |
-----------------------------------------------------+
* = FE nested within cluster; treated as redundant for DoF computation

. di e(r2)
.53325215

.

[Michael Wharff]

Andrew,

This works great. Thank you! As always, you've been a big help.