Dear all,
I am having troubles understanding the reasons why I get different results for my interactions in a pooled and the two unpooled models.
This is an example of my stacked dataset with two obs for each x7.
The following is the model, output, and marginsplot of my pooled analysis:
Unfortunately, when I run the analyses for the two values of x6 I am interested in, the results do not add up. Both are not significant, and while the model for x6=0 is in the expected direction, the p-value is barely <0.10. How can these results be explained?
Sincerely
Mattia
I am having troubles understanding the reasons why I get different results for my interactions in a pooled and the two unpooled models.
This is an example of my stacked dataset with two obs for each x7.
Code:
input float(x1 x2) double(x3 x4) float x5 byte x6 str20 x7 float y
2.5 4 6.1 0 21 2 "AGALEV" 4.818182
1 4 6.1 0 21 1 "AGALEV" 6.416667
3 4 6.1 0 21 0 "AGALEV" 6.25
2.5 6.636364 8.89 1 21 2 "CDV" 4.7272725
3 6.636364 8.89 1 21 1 "CDV" 5.333333
5.2 6.636364 8.89 1 21 0 "CDV" 6.833333
2.5 3.909091 6.14 0 21 2 "ECOLO" 4.818182
1 3.909091 6.14 0 21 1 "ECOLO" 6.272727
3.4 3.909091 6.14 0 21 0 "ECOLO" 6.25
2 7 7.56 1 21 2 "PRL/MR" 5.090909
1.5 7 7.56 1 21 1 "PRL/MR" 4.4545455
2.4 7 7.56 1 21 0 "PRL/MR" 8.083333
2.75 7.5 9.46 0 21 2 "PSBE" 4.2727275
2.5 7.5 9.46 0 21 1 "PSBE" 4
2.2 7.5 9.46 0 21 0 "PSBE" 8.166667
2.3333333 6.7 3.7 0 21 2 "PSC/CDH" 4.3636365
3 6.7 3.7 0 21 1 "PSC/CDH" 5
4.4 6.7 3.7 0 21 0 "PSC/CDH" 6.416667
2.666667 8.125 8.62 0 10 2 "PVDA-PTB" 4
3 8.125 8.62 0 10 1 "PVDA-PTB" 3.4
.6 8.125 8.62 0 10 0 "PVDA-PTB" 8.833333
2.5 6.818182 6.71 0 21 2 "SP/SPA" 4.181818
2.5 6.818182 6.71 0 21 1 "SP/SPA" 4.5833335
2.2 6.818182 6.71 0 21 0 "SP/SPA" 7.916667
2.75 9.2 11.95 0 21 2 "VB" 4.7272725
2 9.2 11.95 0 21 1 "VB" 6.916667
3.6 9.2 11.95 0 21 0 "VB" 4
1.5 6.636364 8.54 1 21 2 "VLD/PVV" 5.090909
1.5 6.636364 8.54 1 21 1 "VLD/PVV" 5.166667
2.6 6.636364 8.54 1 21 0 "VLD/PVV" 8.166667
3.75 8.090909 16.03 0 21 2 "VU/NVA" 4.2727275
5 8.090909 16.03 0 21 1 "VU/NVA" 5.416667
2.8 8.090909 16.03 0 21 0 "VU/NVA" 7.5
2.5714285 4.076923 3 0 1 2 "A" 4.285714
.6666667 4.076923 3 0 1 1 "A" 7.857143
3 4.076923 3 0 1 0 "A" 3.4285715
1.4444444 9.357142 8.7 0 21 2 "DF" 6.785714
1.3333334 9.357142 8.7 0 21 1 "DF" 7.928571
6 9.357142 8.7 0 21 0 "DF" 4.142857
4 3.857143 6.9 0 21 2 "ELDK" 5.769231
1.6666666 3.857143 6.9 0 21 1 "ELDK" 7.142857
1 3.857143 6.9 0 21 0 "ELDK" 7.357143
3.333333 7.285714 6.6 0 21 2 "KF" 4.714286
2.3333333 7.285714 6.6 0 21 1 "KF" 5.642857
1.6666666 7.285714 6.6 0 21 0 "KF" 7.357143
3.875 8.571428 2.3 0 10 2 "LA" 3.642857
5.333333 8.571428 2.3 0 10 1 "LA" 4.428571
1 8.571428 2.3 0 10 0 "LA" 8.428572
2 9.214286 2.4 0 1 2 "NB" 4.714286
1 9.214286 2.4 0 1 1 "NB" 7.583333
2 9.214286 2.4 0 1 0 "NB" 5.714286
1.1111112 7.071429 8.6 0 21 2 "RV" 7.142857
1 7.071429 8.6 0 21 1 "RV" 7.928571
3.333333 7.071429 8.6 0 21 0 "RV" 6
3.111111 6.928571 25.9 1 21 2 "SDDK" 4.714286
3 6.928571 25.9 1 21 1 "SDDK" 5.5
2.3333333 6.928571 25.9 1 21 0 "SDDK" 7.357143
3.888889 4.857143 7.7 0 21 2 "SFDK" 4.571429
1.6666666 4.857143 7.7 0 21 1 "SFDK" 5.571429
1.6666666 4.857143 7.7 0 21 0 "SFDK" 6.142857
4.3333335 6.785714 23.4 0 21 2 "v" 5.214286
6 6.785714 23.4 0 21 1 "v" 5.5
2.666667 6.785714 23.4 0 21 0 "v" 6.714286
2.909091 7 12.6 0 6 2 "AfD" 6.714286
3.166667 7 12.6 0 6 1 "AfD" 9.428572
6.666667 7 12.6 0 6 0 "AfD" 3.190476
2.636364 8.315789 26.8 1 21 2 "CDUGE" 6.857143
5.666667 8.315789 26.8 1 21 1 "CDUGE" 6.238095
4.5833335 8.315789 26.8 1 21 0 "CDUGE" 6.476191
3.636364 8.315789 6.2 1 21 2 "CSU" 6.6
3.4 8.315789 6.2 1 21 1 "CSU" 7.619048
3.090909 8.315789 6.2 1 21 0 "CSU" 6.523809
. . . 0 6 2 "DieTier" 4
3 . . 0 6 1 "DieTier" 7
2 . . 0 6 0 "DieTier" 2.5
3.727273 8.263158 10.7 0 21 2 "FDP" 5.666667
3.083333 8.263158 10.7 0 21 1 "FDP" 5.047619
1.8333334 8.263158 10.7 0 21 0 "FDP" 8.1
1.4545455 4.842105 8.9 0 21 2 "GRUNEN" 7.333333
1.5 4.842105 8.9 0 21 1 "GRUNEN" 8.476191
3.75 4.842105 8.9 0 21 0 "GRUNEN" 5
5.090909 6.055555 9.2 0 21 2 "LINKE" 4.85
4.3333335 6.055555 9.2 0 21 1 "LINKE" 5.238095
1.9166666 6.055555 9.2 0 21 0 "LINKE" 8.095238
0 1.5 . 0 6 2 "Piraten" 5.25
1.6666666 1.5 . 0 6 1 "Piraten" 8.6
2.666667 1.5 . 0 6 0 "Piraten" 2.2
2.3636363 5.578948 20.5 1 21 2 "SPD" 6.857143
4.1666665 5.578948 20.5 1 21 1 "SPD" 5.476191
5.5 5.578948 20.5 1 21 0 "SPD" 7.666667
1.4 9.666667 3.7 0 1 2 "EL" 4.375
1 9.666667 3.7 0 1 1 "EL" 8.25
2.666667 9.666667 3.7 0 1 0 "EL" 4.375
1.5 6.666667 .74 0 1 2 "KIDISO" 7.142857
2 6.666667 .74 0 1 1 "KIDISO" 6.75
2 6.666667 .74 0 1 0 "KIDISO" 7.2
.16666667 9 5.3 0 21 2 "KKE" 6.222222
.5 9 5.3 0 21 1 "KKE" 3.25
0 9 5.3 0 21 0 "KKE" 9.111111
1 5.5 3.44 0 1 2 "MR25" 7.625
end
Code:
eststo seven: reg y c.x1##c.x2 x3 x4 x5 i.country if x6!=2 & in_model_1==1, vce(cl x7)
margins, dydx(x1) at(x2=(0(0.5)10))
marginsplot, title("Average Marginal Effects of x1 on y (95% CIs)") xtitle("x2") ///
addplot(histogram x2 if x6!=2, freq width(0.5) yaxis(2) yscale(alt axis(2)) fcolor(%25) lc(black%50))
Code:
Linear regression Number of obs = 258
F(20, 129) = 11.74
Prob > F = 0.0000
R-squared = 0.3411
Root MSE = 1.2825
(Std. err. adjusted for 130 clusters in x7)
------------------------------------------------------------------------------
| Robust
y | Coefficient std. err. t P>|t| [95% conf. interval]
-------------+----------------------------------------------------------------
x1 | .0019245 .2214274 0.01 0.993 -.436175 .4400241
x2 | .0689307 .0806277 0.85 0.394 -.0905932 .2284545
|
c.x1#c.x2 | -.0763485 .0301067 -2.54 0.012 -.1359155 -.0167816
|
x3 | .0439528 .0079439 5.53 0.000 .0282355 .05967
x4 | .0707958 .146916 0.48 0.631 -.2198811 .3614728
x5 | .0015976 .0101892 0.16 0.876 -.018562 .0217572
|
country |
2. dk | .1339352 .2632911 0.51 0.612 -.3869927 .654863
3. ge | .8146273 .2339428 3.48 0.001 .3517659 1.277489
4. gr | .5158083 .3239174 1.59 0.114 -.1250702 1.156687
5. esp | .0285887 .2713843 0.11 0.916 -.5083519 .5655292
6. fr | .7092217 .2187707 3.24 0.002 .2763785 1.142065
7. irl | -.68028 .2382757 -2.86 0.005 -1.151714 -.2088458
8. it | .2901257 .4193663 0.69 0.490 -.5396007 1.119852
10. nl | -.420152 .1813169 -2.32 0.022 -.7788919 -.0614122
11. uk | -.787641 .2225044 -3.54 0.001 -1.227871 -.3474106
12. por | .7449562 .3259486 2.29 0.024 .1000589 1.389854
13. aus | -.8840575 .3138242 -2.82 0.006 -1.504966 -.2631485
14. fin | .6545474 .1682645 3.89 0.000 .3216321 .9874628
16. sv | .4814474 .216804 2.22 0.028 .0524953 .9103995
38. lux | -1.392723 .4609494 -3.02 0.003 -2.304723 -.4807235
|
_cons | 6.653327 .6476639 10.27 0.000 5.371908 7.934746
------------------------------------------------------------------------------
Unfortunately, when I run the analyses for the two values of x6 I am interested in, the results do not add up. Both are not significant, and while the model for x6=0 is in the expected direction, the p-value is barely <0.10. How can these results be explained?
Code:
eststo seven2: reg y c.x1##c.x2 x3 x4 x5 i.country if x6==0 & in_model_2==1
margins, dydx(x1) at(x2=(0(0.5)10))
marginsplot, title("Average Marginal Effects of x1 on y (95% CIs)") xtitle("x2") ///
addplot(histogram x2 if x6==0 & in_model_2==1, freq width(0.5) yaxis(2) yscale(alt axis(2)) fcolor(%25) lc(black%50))
eststo seven3: reg y c.x1##c.x2 x3 x4 x5 i.country if x6==1 & in_model_3==1
margins, dydx(x1) at(x2=(0(0.5)10))
marginsplot, title("Average Marginal Effects of x1 on y (95% CIs)") xtitle("x2") ///
addplot(histogram x2 if x6==1 & in_model_3==1, freq width(0.5) yaxis(2) yscale(alt axis(2)) fcolor(%25) lc(black%50))
Code:
eststo seven2: reg y c.x1##c.x2 x3 x4 x5 i.country if x6==0 & in_model_2==1
Source | SS df MS Number of obs = 128
-------------+---------------------------------- F(20, 107) = 4.76
Model | 150.651879 20 7.53259395 Prob > F = 0.0000
Residual | 169.297938 107 1.58222372 R-squared = 0.4709
-------------+---------------------------------- Adj R-squared = 0.3720
Total | 319.949817 127 2.5192899 Root MSE = 1.2579
------------------------------------------------------------------------------
y | Coefficient Std. err. t P>|t| [95% conf. interval]
-------------+----------------------------------------------------------------
x1 | .1074972 .3360307 0.32 0.750 -.5586445 .7736388
x2 | .0709244 .180704 0.39 0.695 -.2873001 .429149
|
c.x1#c.x2 | -.08677 .0495838 -1.75 0.083 -.1850642 .0115242
|
x3 | .0609723 .016123 3.78 0.000 .0290104 .0929342
x4 | .137091 .3109673 0.44 0.660 -.4793655 .7535475
x5 | .0312748 .0174638 1.79 0.076 -.0033452 .0658947
|
country |
2. dk | -.8865927 .5647997 -1.57 0.119 -2.006242 .2330568
3. ge | -.4909954 .6170599 -0.80 0.428 -1.714245 .7322539
4. gr | .0128348 .6165485 0.02 0.983 -1.209401 1.23507
5. esp | -.3776846 .531973 -0.71 0.479 -1.432259 .6768899
6. fr | .3147261 .5968196 0.53 0.599 -.868399 1.497851
7. irl | -1.004265 .6580029 -1.53 0.130 -2.308679 .3001494
8. it | -.3107438 .6299975 -0.49 0.623 -1.55964 .9381529
10. nl | -1.235017 .5238642 -2.36 0.020 -2.273517 -.1965178
11. uk | -2.06065 .6022026 -3.42 0.001 -3.254446 -.8668532
12. por | .5106341 .6342881 0.81 0.423 -.7467681 1.768036
13. aus | -1.83555 .699734 -2.62 0.010 -3.222691 -.4484088
14. fin | -.5542002 .6047479 -0.92 0.362 -1.753042 .6446421
16. sv | -.7864335 .5936533 -1.32 0.188 -1.963282 .390415
38. lux | -1.986957 .6895806 -2.88 0.005 -3.35397 -.6199439
|
_cons | 6.848582 1.376672 4.97 0.000 4.11949 9.577673
------------------------------------------------------------------------------
eststo seven3: reg y c.x1##c.x2 x3 x4 x5 i.country if x6==1 & in_model_3==1
Source | SS df MS Number of obs = 130
-------------+---------------------------------- F(20, 109) = 5.44
Model | 132.957861 20 6.64789305 Prob > F = 0.0000
Residual | 133.32071 109 1.22312578 R-squared = 0.4993
-------------+---------------------------------- Adj R-squared = 0.4075
Total | 266.278571 129 2.06417497 Root MSE = 1.106
------------------------------------------------------------------------------
y | Coefficient Std. err. t P>|t| [95% conf. interval]
-------------+----------------------------------------------------------------
x1 | -.6511678 .3113023 -2.09 0.039 -1.268159 -.0341768
x2 | -.0603839 .112382 -0.54 0.592 -.2831215 .1623536
|
c.x1#c.x2 | .0043718 .0439778 0.10 0.921 -.0827908 .0915343
|
x3 | .0330859 .0147282 2.25 0.027 .0038952 .0622767
x4 | -.0551796 .2737978 -0.20 0.841 -.5978379 .4874787
x5 | -.0272743 .014766 -1.85 0.067 -.05654 .0019914
|
country |
2. dk | 1.199387 .4903972 2.45 0.016 .2274362 2.171339
3. ge | 2.207378 .54445 4.05 0.000 1.128296 3.28646
4. gr | 1.114636 .5351736 2.08 0.040 .0539394 2.175333
5. esp | .4695195 .4680648 1.00 0.318 -.4581697 1.397209
6. fr | 1.346269 .5183206 2.60 0.011 .3189745 2.373564
7. irl | -.153926 .5391264 -0.29 0.776 -1.222457 .914605
8. it | 1.005785 .5587891 1.80 0.075 -.1017168 2.113287
10. nl | .46 .4622662 1.00 0.322 -.4561965 1.376197
11. uk | .6379496 .5302325 1.20 0.232 -.4129541 1.688853
12. por | 1.022597 .5574282 1.83 0.069 -.0822071 2.127402
13. aus | -.0046027 .6292805 -0.01 0.994 -1.251816 1.242611
14. fin | 2.056141 .5390641 3.81 0.000 .9877332 3.124548
16. sv | 1.785488 .5192244 3.44 0.001 .756402 2.814574
38. lux | -.5289811 .6191384 -0.85 0.395 -1.756093 .698131
|
_cons | 7.331186 .8890873 8.25 0.000 5.569044 9.093328
------------------------------------------------------------------------------
Sincerely
Mattia
Comment