Hi Everyone!
For my master's thesis, I am conducting regression estimation using the Two-Way Fixed Effect method. My dataset comprises six rounds of panel data from 10,802 communities. The aim is to test how Humanitarian Aid (binary variable) delivered to communities attracts Internally Displaced Persons ArrivalIDPs (continuous variable) compared to those that have not received the Aid.
Start with the basic estimation:
Code:
The Result: 
#1: Why is R-squared not available for the random effect? I am aware that I can find this separately adj_R^2 by adding r2_b, r2_w, and without r2_a, but just wondering if it is still possible to include it directly in the output. Also, the R-squared is extremely low.
#2: After conducting the Hausman Test, it confirmed that the fixed effect (xtreg, fe) is the best fit. However, even with adding many control variables, the R-squared still remains extremely low. Despite this, the coefficients are significant, especially for the extreme categories within the categorical variable (e.g., =5), which is what I am looking for and makes sense. What could be the issue and the possible solution? I have uploaded the descriptions of the variables below if needed.
Result:
Will appreciate any insights and assistance.
For my master's thesis, I am conducting regression estimation using the Two-Way Fixed Effect method. My dataset comprises six rounds of panel data from 10,802 communities. The aim is to test how Humanitarian Aid (binary variable) delivered to communities attracts Internally Displaced Persons ArrivalIDPs (continuous variable) compared to those that have not received the Aid.
Start with the basic estimation:
Code:
Code:
xtset SettlementID Round reg ArrivalIDPs HUMDelivered outreg2 using my_reg1.doc, replace ctitle(Poold OLS) /// keep(HUMDelivered ) /// addtext(Settlement FE, YES) xtreg ArrivalIDPs HUMDelivered, re outreg2 using my_reg1.doc, append ctitle(Random-Effect) /// keep(HUMDelivered ) /// addtext(Settlement FE, YES) xtreg ArrivalIDPs HUMDelivered, fe outreg2 using my_reg1.doc, append ctitle(Fixed-Effect) /// keep(HUMDelivered i.Round ) /// addtext(Settlement FE, YES) reg ArrivalIDPs HUMDelivered i.Round outreg2 using my_reg1.doc, append ctitle(Poold OLS) /// keep(HUMDelivered i.Round) /// addtext(Year FE, YES, Settlement FE, YES) xtreg ArrivalIDPs HUMDelivered i.Round , re outreg2 using my_reg1.doc, append ctitle(Random-Effect) /// keep(HUMDelivered i.Round) /// addtext(Year FE, YES, Settlement FE, YES) xtreg ArrivalIDPs HUMDelivered i.Round , fe outreg2 using my_reg1.doc, append ctitle(Fixed-Effect) /// keep(HUMDelivered i.Round) /// addtext(Year FE, YES, Settlement FE, YES)
#1: Why is R-squared not available for the random effect? I am aware that I can find this separately adj_R^2 by adding r2_b, r2_w, and without r2_a, but just wondering if it is still possible to include it directly in the output. Also, the R-squared is extremely low.
#2: After conducting the Hausman Test, it confirmed that the fixed effect (xtreg, fe) is the best fit. However, even with adding many control variables, the R-squared still remains extremely low. Despite this, the coefficients are significant, especially for the extreme categories within the categorical variable (e.g., =5), which is what I am looking for and makes sense. What could be the issue and the possible solution? I have uploaded the descriptions of the variables below if needed.
Result:
Code:
. xtreg ArrivalIDPs HUMDelivered i.Round i.IDPInConflicts i.IDPsNatDisaster i.pashtun_greg i.HLTClinics i.EduSchoolExist, fe
note: 1.pashtun_greg omitted because of collinearity.
Fixed-effects (within) regression Number of obs = 64,620
Group variable: SettlementID Number of groups = 10,770
R-squared: Obs per group:
Within = 0.0032 min = 6
Between = 0.0445 avg = 6.0
Overall = 0.0166 max = 6
F(18, 53832) = 9.69
corr(u_i, Xb) = 0.1131 Prob > F = 0.0000
----------------------------------------------------------------------------------
ArrivalIDPs | Coefficient Std. err. t P>|t| [95% conf. interval]
-----------------+----------------------------------------------------------------
HUMDelivered | 11.23654 6.46464 1.74 0.082 -1.434206 23.90729
|
Round |
11 | 2.771986 8.585365 0.32 0.747 -14.0554 19.59937
12 | -10.64786 8.612658 -1.24 0.216 -27.52874 6.233019
13 | 15.06621 8.631937 1.75 0.081 -1.852459 31.98487
14 | 52.79905 8.635658 6.11 0.000 35.87309 69.72501
16 | 31.35202 9.322178 3.36 0.001 13.08047 49.62356
|
IDPInConflicts |
1 | -19.43057 13.52439 -1.44 0.151 -45.93847 7.077339
2 | -3.425755 12.49668 -0.27 0.784 -27.91935 21.06784
3 | -1.334446 11.44944 -0.12 0.907 -23.77544 21.10655
4 | 1.823892 11.74439 0.16 0.877 -21.1952 24.84298
5 | 31.98156 10.9981 2.91 0.004 10.42519 53.53793
|
IDPsNatDisaster |
1 | 30.34869 12.99162 2.34 0.019 4.885003 55.81237
2 | 8.353065 11.92255 0.70 0.484 -15.01522 31.72135
3 | 35.46722 10.80746 3.28 0.001 14.2845 56.64994
4 | 41.21825 11.11009 3.71 0.000 19.44237 62.99412
5 | 70.43386 11.74086 6.00 0.000 47.42169 93.44604
|
1.pashtun_greg | 0 (omitted)
1.HLTClinics | -11.40218 10.1488 -1.12 0.261 -31.2939 8.489543
1.EduSchoolExist | 13.95478 8.30587 1.68 0.093 -2.324793 30.23435
_cons | 353.908 9.975406 35.48 0.000 334.3562 373.4599
-----------------+----------------------------------------------------------------
sigma_u | 1751.7956
sigma_e | 627.38448
rho | .88631833 (fraction of variance due to u_i)
----------------------------------------------------------------------------------
F test that all u_i=0: F(10769, 53832) = 45.58 Prob > F = 0.0000
Will appreciate any insights and assistance.
Comment