[R] Confusion going from Stata -> R
Rob Denniker
bearmarketsrule at inbox.com
Fri May 1 04:50:27 CEST 2009
Dear list,
I am trying to replicate some Stata results but having a tough time doing it in R. The goal is to obtain a difference-in-difference estimate in a model with simple state fixed effects. The "state" variable is a factor, but some levels are missing. It appears that Stata automatically recognizes this and works around it. It also automatically pick "ME" as the base state (see below), but trying to do the same thing in R does not yield the same results...
The data is as follows:
# States were assigned to treatment and control groups in 1976, and residents surveyed in 1976 and 1985
# i.year indicates the year the respondent was surveyed
# i.st is a factor indicating the respondent's state of residence
# i.treatment is a dummy indicating whether that individual was resident in a treatment state
# hwover is the binary response variable of interest
load(url("http://h1.ripway.com/rdump/d.rdata"))
# load data frame "d" (slow to load, retry if 404)
# Code i.st as a factor
d$i.st <- factor(d$st)
n.i.st <- length(levels(d$i.st))
levels(d$i.st)[14] # Level 14 == "ME"
contrasts(d$i.st) <- contr.treatment(n.i.st, base = 14, contrasts=T) # Sets base level to "ME"
attributes(d$i.st)
# Difference-in-difference estimate
reg1 <- lm(hwover ~ i.treatment*i.year + i.st, data=d, weights=weight1)
summary(reg1)
# Output (correctly) omit state "14" ("ME")
# Estimates of average treatment effect (coefficient on i.treatment1:i.year1985) same as Stata's
# So is the coefficient on i.year1985.
# All other params are off.
For comparison, I am pasting Stata's output below.
Any ideas would be much appreciated!
xi: reg hwover i.treatment*i.year i.st if st1976==1 [pw=weight1]
i.treatment _Itreatment_0-1 (naturally coded; _Itreatment_0 omitted)
(characteristic year[omit]: 1968
yet variable year never equals 1968; characteristic ignored)
i.year _Iyear_1976-1985 (naturally coded; _Iyear_1976 omitted)
i.tre~t*i.year _ItreXyea_#_# (coded as above)
i.st _Ist_1-52 (_Ist_23 for st==ME omitted)
(sum of wgt is 6.0268e+03)
Linear regression Number of obs = 5976
F( 37, 5938) = 1.37
Prob > F = 0.0650
R-squared = 0.0083
Root MSE = .32063
------------------------------------------------------------------------------
| Robust
hwover | Coef. Std. Err. t P>|t| [95% Conf. Interval]
-------------+----------------------------------------------------------------
_Itreatmen~1 | .0117852 .0438655 0.27 0.788 -.0742072 .0977775
_Iyear_1985 | .0297313 .0163587 1.82 0.069 -.0023377 .0618004
_ItreXy~1985 | -.055271 .0198304 -2.79 0.005 -.0941457 -.0163963
_Ist_1 | (dropped)
_Ist_2 | (dropped)
_Ist_3 | -.0029799 .0325302 -0.09 0.927 -.066751 .0607912
_Ist_4 | (dropped)
_Ist_5 | .0472988 .0408782 1.16 0.247 -.0328373 .1274349
_Ist_6 | .072884 .0292795 2.49 0.013 .0154855 .1302826
_Ist_7 | .0995722 .0440573 2.26 0.024 .0132039 .1859406
_Ist_8 | .0489226 .0424768 1.15 0.249 -.0343473 .1321926
_Ist_9 | (dropped)
_Ist_10 | (dropped)
_Ist_11 | .0591339 .0341094 1.73 0.083 -.0077329 .1260008
_Ist_12 | .0229454 .0355917 0.64 0.519 -.0468274 .0927181
_Ist_13 | (dropped)
_Ist_14 | (dropped)
_Ist_15 | .0058994 .0511652 0.12 0.908 -.094403 .1062018
_Ist_16 | .0830382 .0326732 2.54 0.011 .0189868 .1470896
_Ist_17 | .0579801 .0385756 1.50 0.133 -.0176421 .1336022
_Ist_18 | -.0013615 .0342379 -0.04 0.968 -.0684803 .0657573
_Ist_19 | (dropped)
_Ist_20 | .0310298 .0428693 0.72 0.469 -.0530095 .1150692
_Ist_21 | (dropped)
_Ist_22 | .0302868 .0448914 0.67 0.500 -.0577166 .1182902
_Ist_24 | .060111 .0334748 1.80 0.073 -.0055119 .1257338
_Ist_25 | .0003036 .0329512 0.01 0.993 -.0642926 .0648999
_Ist_26 | -.00681 .0307812 -0.22 0.825 -.0671522 .0535323
_Ist_27 | .0452638 .0503703 0.90 0.369 -.0534803 .1440078
_Ist_28 | .0869412 .0867014 1.00 0.316 -.083025 .2569075
_Ist_29 | -.0059285 .0384747 -0.15 0.878 -.0813529 .0694959
_Ist_30 | (dropped)
_Ist_31 | .1421038 .0558453 2.54 0.011 .0326267 .2515808
_Ist_32 | (dropped)
_Ist_33 | .0552327 .0331347 1.67 0.096 -.0097233 .1201887
_Ist_34 | (dropped)
_Ist_35 | (dropped)
_Ist_36 | .0245104 .0377347 0.65 0.516 -.0494633 .0984841
_Ist_37 | .0558682 .0307152 1.82 0.069 -.0043448 .1160812
_Ist_38 | (dropped)
_Ist_39 | .0429489 .04253 1.01 0.313 -.0404253 .1263232
_Ist_40 | .0230925 .0301413 0.77 0.444 -.0359955 .0821804
_Ist_41 | (dropped)
_Ist_42 | .0408904 .0374549 1.09 0.275 -.0325349 .1143157
_Ist_43 | (dropped)
_Ist_44 | .0039713 .0431761 0.09 0.927 -.0806695 .0886122
_Ist_45 | .0425491 .0295203 1.44 0.150 -.0153213 .1004196
_Ist_46 | .0148853 .0447929 0.33 0.740 -.0729251 .1026957
_Ist_47 | .0146356 .0427135 0.34 0.732 -.0690984 .0983695
_Ist_48 | .1538798 .0817703 1.88 0.060 -.0064197 .3141793
_Ist_49 | .0191617 .0332331 0.58 0.564 -.0459873 .0843107
_Ist_50 | .033731 .0389537 0.87 0.387 -.0426325 .1100945
_Ist_51 | .0431437 .0482243 0.89 0.371 -.0513934 .1376808
_Ist_52 | (dropped)
_cons | .0765964 .0365322 2.10 0.036 .00498 .1482128
------------------------------------------------------------------------------
More information about the R-help
mailing list