Statalist The Stata Listserver


[Date Prev][Date Next][Thread Prev][Thread Next][Date index][Thread index]

Re: st: Pscore and balancing


From   "Gauri Khanna" <[email protected]>
To   [email protected]
Subject   Re: st: Pscore and balancing
Date   Thu, 12 Apr 2007 12:40:41 +0000

Dear Stata List,

As a follow up to my last email, I am still grappling with the balancing property using the pscore module. Further to Ricardo's suggestion, I used fewer variables to estimate the propensity score. However, no matter what combination I use, I still get the error. I wonder if any one else has has this problem and what they did to overcome it. I will grateful for your help.

My command is :

-pscore waterpipe v151 v152 v012 ageresp2 aware ///
v512 v715 eduhusb eduhusb2 v730 partnerage2 v025 ///
sh42 , pscore(mypscore) blockid(myblock) comsup numblo(5) level(0.005) logit-

where my variables are

v151 sex of household head
v152 age of household head
v012 current age - respondent
ageresp2 square of current age-respondent
aware awareness (listens to radio/ watches tv/ readsnewspaper)
v512 years since first marriage
v715 husbands education-single yrs
eduhusb square of husbands education-single yrs
eduhusb2 cube of husbands education-single yrs
v730 partners age
partnerage2 sqaure of partmers age
v025 type of place of residence
sh42 does this household own this
house or any other house?

The output is :

****************************************************
Algorithm to estimate the propensity score
****************************************************


The treatment is waterpipe

waterpipe | Freq. Percent Cum.
------------+-----------------------------------
0 | 15,396 61.91 61.91
1 | 9,473 38.09 100.00
------------+-----------------------------------
Total | 24,869 100.00



Estimation of the propensity score

Iteration 0: log likelihood = -16525.718
Iteration 1: log likelihood = -13670.121
Iteration 2: log likelihood = -13630.618
Iteration 3: log likelihood = -13630.359
Iteration 4: log likelihood = -13630.359

Logistic regression Number of obs = 24869
LR chi2(13) = 5790.72
Prob > chi2 = 0.0000
Log likelihood = -13630.359 Pseudo R2 = 0.1752

------------------------------------------------------------------------------
waterpipe | Coef. Std. Err. z P>|z| [95% Conf. Interval]
-------------+----------------------------------------------------------------
v151 | -.0863459 .0610051 -1.42 0.157 -.2059138 .033222
v152 | -.0029664 .0010262 -2.89 0.004 -.0049776 -.0009551
v012 | .0485913 .0219485 2.21 0.027 .005573 .0916097
ageresp2 | .0000214 .0003888 0.06 0.956 -.0007406 .0007833
aware | .7133884 .0343954 20.74 0.000 .6459746 .7808022
v512 | -.0718782 .0052339 -13.73 0.000 -.0821364 -.06162
v715 | .0415108 .0208768 1.99 0.047 .000593 .0824286
eduhusb | -.0047051 .0033843 -1.39 0.164 -.0113382 .001928
eduhusb2 | .000216 .0001408 1.53 0.125 -.00006 .000492
v730 | .0366385 .010276 3.57 0.000 .016498 .0567791
partnerage2 | -.0004604 .0001197 -3.84 0.000 -.000695 -.0002257
v025 | -1.659859 .0351148 -47.27 0.000 -1.728683 -1.591036
sh42 | -.2596864 .0528507 -4.91 0.000 -.3632719 -.156101
_cons | .8748132 .3151323 2.78 0.006 .2571653 1.492461
------------------------------------------------------------------------------



Note: the common support option has been selected
The region of common support is [.06287784, .90866427]



Description of the estimated propensity score
in region of common support

Estimated propensity score
-------------------------------------------------------------
Percentiles Smallest
1% .120273 .0628778
5% .1411216 .063518
10% .1534986 .0638564 Obs 24823
25% .1842821 .0689311 Sum of Wgt. 24823

50% .3147234 Mean .3815467
Largest Std. Dev. .2302978
75% .5411638 .9032886
90% .7707648 .9035941 Variance .0530371
95% .8032801 .9039739 Skewness .7839778
99% .851611 .9086643 Kurtosis 2.176904



******************************************************
Step 1: Identification of the optimal number of blocks
Use option detail if you want more detailed output
******************************************************


The final number of blocks is 13

This number of blocks ensures that the mean propensity score
is not different for treated and controls in each blocks



**********************************************************
Step 2: Test of balancing property of the propensity score
Use option detail if you want more detailed output
**********************************************************

Variable v730 is not balanced in block 9

Variable eduhusb is not balanced in block 12

Variable eduhusb2 is not balanced in block 12

Variable v715 is not balanced in block 13

Variable eduhusb is not balanced in block 13

Variable eduhusb2 is not balanced in block 13

The balancing property is not satisfied

Try a different specification of the propensity score

Inferior |
of block | waterpipe
of pscore | 0 1 | Total
-----------+----------------------+----------
0 | 45 3 | 48
.1 | 1,811 247 | 2,058
.15 | 2,560 461 | 3,021
.175 | 1,998 485 | 2,483
.2 | 1,659 479 | 2,138
.25 | 1,348 497 | 1,845
.3 | 1,955 947 | 2,902
.35 | 1,379 878 | 2,257
.4 | 847 692 | 1,539
.5 | 306 417 | 723
.6 | 318 678 | 996
.7 | 849 2,631 | 3,480
.8 | 275 1,058 | 1,333
-----------+----------------------+----------
Total | 15,350 9,473 | 24,823

Note: the common support option has been selected


*******************************************
End of the algorithm to estimate the pscore
*******************************************

Sincerely,

Gauri


From: "Gauri Khanna" <[email protected]>
Reply-To: [email protected]
To: [email protected]
Subject: Re: st: Pscore and balancing
Date: Wed, 11 Apr 2007 16:45:11 +0000

Dear Ricardo,

Thank you for your prompt response:

1. So that I know what you mean by "more parsimonious", you are suggesting to start off with fewer variables than my current set of 73 to generate the pscore? Am I right?

The reason I say this is because in Becker and Ichino's article in Stata Journal, Vol 2, No 4 the exact opposite is stated... if the balancing property is not satisfied a "less parsimonious" specification is needed.

2. A related question, if I start off with fewer variables is there a criteria that one uses to pick and choose?

(To be on the safe side I looked up the dictionary meaning of parsimonious and it means frugal.)


I repeat the other two questions that perhaps you or anyone else might know

1. Firstly, why does the number of observations get reduced to

13821 from 24,672 in the logit estimation?
4. browsed at the pscore and blockid variables generated "pscorepump" and "myblock" and found pscorepump==. for myblock==0. I don't understand what this means. There are atleast 10,000 missing values for pscorepump.

I appreciate your continued advice.


Regards,

Gauri

PS. Thanks for the reference of Deheja and Wahba (i had their working paper of this and I will see the final version also)



From: "ricardo sierra" <[email protected]>
Reply-To: [email protected]
To: [email protected]
Subject: Re: st: Pscore and balancing
Date: Wed, 11 Apr 2007 11:39:29 -0400

Is there a way you can start with a more parsimonious logit
specification to estimate de score ??

** This is the 1st step proposed in Dehejia and Wahba (2002)


On 4/11/07, Gauri Khanna <[email protected]> wrote:
Dear Stata List,

I am using Stata Version 9.2. I am generating propensity scores using the
pscore module (authors Becker & Ichino 2002) and it is updated. I am stuck
with the balancing property which remains unsatisfied. I have 73 variables,
mostly categorical or dummy, for generating pscore. I have copied the output
below for reference. I have also copied a description of my variables, in
case you would like to scan through them. I looked at the statalist archives
but did not find a response to my question.

This is the command that I use: "waterpump, a binary variable, is my
treatment"

-pscore waterpump v136 v151 v152 hv217 v012 v109 v110 v112 v115 v119 sh39
v131 v137 ///
v208 v218 v404 v416 husbwifelive v512 v627 v628 v629 ///
v715 v705 v717 v718 v714 v731 v730 b9 s509 s510 s511a s511b s511d s512a ///
s512b s513 sh29 sh32a sh32b sh32c sh32d sh32e sh32f sh32x sh34 sh35 ///
sh36 sh37 sh42 sh43 sh46 sh49 sh47a sh47b sh47c sh47d sh47e sh47f ///
sh47g sh47h sh47j sh47k sh47l sh47p sh47q sh47r sh47s sh47t ownerTV ///
hv211 v024, pscore(pscorepump) blockid(myblock) numblo(5) level(0.005)
logit-

1. Firstly, why does the number of observations get reduced to 13821 from
24,672 in the logit estimation?
(When I used the -comsup-option, the no. of observations in the table
"inferior of block pscore" then the no. of observations gets reduced to
13000 odd from 24,762 which is expected. But in the output below, I have not
run -comsup- so the no. of observations remain 24,762)

2. Is there an efficienty way of trying to meet the balancing property?
Since I have so many variables this might be a near impossible task.

3. I understand that one of the ways to meet the balancing property is to
use higher order terms and interactions. Given that I have many dummy
variables or categorical variables this suggestion is not very useful.

4. I browsed at the pscore and blockid variables generated "pscorepump" and
"myblock" and found pscorepump==. for myblock==0. I don't understand what
this means. There are atleast 10,000 missing values for pscorepump.

Given that I have generate propensity scores several times over for
different treatment groups I would be extremely grateful if someone could
please advise me.

Sincerely,

Gauri

****************************************************
Algorithm to estimate the propensity score
****************************************************


The treatment is waterpump

waterpump | Freq. Percent Cum.
------------+-----------------------------------
0 | 15,885 64.15 64.15
1 | 8,877 35.85 100.00
------------+-----------------------------------
Total | 24,762 100.00



Estimation of the propensity score

Iteration 0: log likelihood = -9109.675
Iteration 1: log likelihood = -8131.2148
Iteration 2: log likelihood = -8090.9179
Iteration 3: log likelihood = -8089.9872
Iteration 4: log likelihood = -8089.9861
Iteration 5: log likelihood = -8089.9861

Logistic regression Number of obs =
13821
LR chi2(73) =
2039.38
Prob > chi2 =
0.0000
Log likelihood = -8089.9861 Pseudo R2 =
0.1119

------------------------------------------------------------------------------
waterpump | Coef. Std. Err. z P>|z| [95% Conf.
Interval]
-------------+----------------------------------------------------------------
v136 | .0165368 .0096416 1.72 0.086 -.0023604
.035434
v151 | -.0437896 .0882641 -0.50 0.620 -.2167841
.129205
v152 | .0003202 .0016804 0.19 0.849 -.0029733
.0036138
hv217 | -.0211372 .024753 -0.85 0.393 -.0696522
.0273777
v012 | -.0370832 .0079621 -4.66 0.000 -.0526886
-.0214777
v109 | .0142181 .0726133 0.20 0.845 -.1281014
.1565376
v110 | -.1876761 .0590738 -3.18 0.001 -.3034587
-.0718935
v112 | -.0562225 .0476427 -1.18 0.238 -.1496003
.0371554
v115 | -.0058888 .0009844 -5.98 0.000 -.0078182
-.0039594
v119 | -.9156391 .1534767 -5.97 0.000 -1.216448
-.6148303
sh39 | -.0093317 .0046107 -2.02 0.043 -.0183686
-.0002948
v131 | -.0449353 .017664 -2.54 0.011 -.0795562
-.0103145
v137 | -.0812052 .0256932 -3.16 0.002 -.131563
-.0308473
v208 | .0150714 .0351368 0.43 0.668 -.0537955
.0839383
v218 | -.0535865 .0225205 -2.38 0.017 -.0977259
-.0094471
v404 | .2572917 .0560506 4.59 0.000 .1474345
.3671488
v416 | -.0089503 .0161706 -0.55 0.580 -.0406442
.0227436
husbwifelive | .1250677 .0961814 1.30 0.193 -.0634444
.3135797
v512 | .0439059 .0087173 5.04 0.000 .0268204
.0609914
v627 | -.0036405 .0259638 -0.14 0.888 -.0545286
.0472475
v628 | .0055433 .0290203 0.19 0.849 -.0513354
.062422
v629 | -.0025496 .0170102 -0.15 0.881 -.0358891
.0307898
v715 | .0156505 .0050324 3.11 0.002 .0057872
.0255139
v705 | .0009952 .0026297 0.38 0.705 -.0041589
.0061492
v717 | .0157639 .0075627 2.08 0.037 .0009413
.0305866
v718 | -.054141 .0170187 -3.18 0.001 -.0874971
-.020785
v714 | .5117575 .2638649 1.94 0.052 -.0054081
1.028923
v731 | -.2636456 .1454615 -1.81 0.070 -.5487448
.0214537
v730 | .0071485 .0034344 2.08 0.037 .0004172
.0138798
b9 | -.1381251 .1456387 -0.95 0.343 -.4235717
.1473215
s509 | -.0037555 .0020511 -1.83 0.067 -.0077756
.0002645
s510 | .0039359 .0022827 1.72 0.085 -.0005381
.0084099
s511a | .0046628 .0172106 0.27 0.786 -.0290693
.038395
s511b | .0244404 .0220627 1.11 0.268 -.0188018
.0676825
s511d | -.0662242 .0226893 -2.92 0.004 -.1106944
-.021754
s512a | .0551096 .0352604 1.56 0.118 -.0139995
.1242188
s512b | .1820999 .0517648 3.52 0.000 .0806427
.283557
s513 | -.1124802 .0398112 -2.83 0.005 -.1905087
-.0344516
sh29 | .0133232 .0017694 7.53 0.000 .0098552
.0167912
sh32a | -.0457174 .1929909 -0.24 0.813 -.4239726
.3325379
sh32b | -.4238643 .2675577 -1.58 0.113 -.9482677
.1005391
sh32c | .0914454 .2024985 0.45 0.652 -.3054444
.4883352
sh32d | -.7686991 .1828472 -4.20 0.000 -1.127073
-.4103252
sh32e | -.3027793 .6182508 -0.49 0.624 -1.514529
.9089701
sh32f | .3425503 .1947313 1.76 0.079 -.039116
.7242167
sh32x | .0599032 .3102354 0.19 0.847 -.548147
.6679534
sh34 | -.4306421 .1426025 -3.02 0.003 -.7101378
-.1511464
sh35 | -.0315757 .0168604 -1.87 0.061 -.0646214
.00147
sh36 | -.1901594 .0456417 -4.17 0.000 -.2796155
-.1007033
sh37 | .0214189 .0073974 2.90 0.004 .0069202
.0359176
sh42 | .2409778 .0744005 3.24 0.001 .0951556
.3868001
sh43 | -.1217373 .0462229 -2.63 0.008 -.2123325
-.0311421
sh46 | .02258 .0465304 0.49 0.627 -.0686178
.1137779
sh49 | .0316219 .0306231 1.03 0.302 -.0283983
.0916421
sh47a | -.281183 .0458982 -6.13 0.000 -.3711419
-.1912241
sh47b | -.3105555 .067574 -4.60 0.000 -.4429981
-.1781129
sh47c | -.2774826 .0570406 -4.86 0.000 -.3892801
-.1656852
sh47d | .2857967 .0507951 5.63 0.000 .1862401
.3853534
sh47e | -.0465287 .0600335 -0.78 0.438 -.1641923
.0711348
sh47f | -.0569313 .0474101 -1.20 0.230 -.1498534
.0359908
sh47g | .1779822 .0602837 2.95 0.003 .0598282
.2961362
sh47h | .453847 .0426894 10.63 0.000 .3701772
.5375167
sh47j | -.0186251 .0684948 -0.27 0.786 -.1528723
.1156222
sh47k | .2616956 .2048625 1.28 0.201 -.1398276
.6632188
sh47l | .0638542 .1556463 0.41 0.682 -.241207
.3689154
sh47p | .1524665 .3714589 0.41 0.681 -.5755795
.8805126
sh47q | .4475498 .0971597 4.61 0.000 .2571203
.6379794
sh47r | .382921 .0751051 5.10 0.000 .2357178
.5301243
sh47s | .666872 .1650351 4.04 0.000 .3434092
.9903348
sh47t | -.3674771 .1922123 -1.91 0.056 -.7442063
.0092521
ownerTV | .126107 .0726295 1.74 0.083 -.0162441
.2684582
hv211 | .2028699 .1075005 1.89 0.059 -.0078272
.413567
v024 | .0062517 .0025395 2.46 0.014 .0012744
.0112291
_cons | .0984652 .4290789 0.23 0.818 -.742514
.9394444
------------------------------------------------------------------------------



Description of the estimated propensity score

Estimated propensity score
-------------------------------------------------------------
Percentiles Smallest
1% .0435489 .0014547
5% .0808509 .0104286
10% .1245135 .0129766 Obs 13821
25% .2320692 .0131112 Sum of Wgt. 13821

50% .3675371 Mean .3703061
Largest Std. Dev. .1787873
75% .5033935 .9397132
90% .6070773 .9507915 Variance .0319649
95% .6681031 .9670365 Skewness .1283541
99% .7641454 .9676308 Kurtosis 2.319066



******************************************************
Step 1: Identification of the optimal number of blocks
Use option detail if you want more detailed output
******************************************************


The final number of blocks is 14

This number of blocks ensures that the mean propensity score
is not different for treated and controls in each blocks



**********************************************************
Step 2: Test of balancing property of the propensity score
Use option detail if you want more detailed output
**********************************************************

Variable v717 is not balanced in block 1

Variable sh47h is not balanced in block 1

Variable v115 is not balanced in block 2

Variable v705 is not balanced in block 2

Variable s511a is not balanced in block 2

Variable sh32x is not balanced in block 2

Variable sh43 is not balanced in block 2

Variable sh47r is not balanced in block 2

Variable sh42 is not balanced in block 3

Variable v119 is not balanced in block 4

Variable sh32a is not balanced in block 4

Variable sh34 is not balanced in block 4

Variable sh47d is not balanced in block 4

Variable sh47g is not balanced in block 4

Variable sh32f is not balanced in block 5

Variable sh32a is not balanced in block 6

Variable sh32f is not balanced in block 6

Variable sh47h is not balanced in block 6

Variable sh32a is not balanced in block 9

Variable sh32f is not balanced in block 9

Variable sh47g is not balanced in block 11

Variable sh47j is not balanced in block 11

Variable v115 is not balanced in block 12

Variable sh47q is not balanced in block 12

Variable s512a is not balanced in block 13

Variable s512b is not balanced in block 13

Variable sh29 is not balanced in block 13

Variable sh47c is not balanced in block 13

The balancing property is not satisfied

Try a different specification of the propensity score

Inferior |
of block | waterpump
of pscore | 0 1 | Total
-----------+----------------------+----------
0 | 7,398 3,765 | 11,163
.05 | 370 18 | 388
.075 | 340 34 | 374
.1 | 696 107 | 803
.15 | 754 174 | 928
.2 | 912 251 | 1,163
.25 | 895 339 | 1,234
.3 | 1,745 890 | 2,635
.4 | 1,383 1,160 | 2,543
.5 | 517 610 | 1,127
.55 | 359 546 | 905
.6 | 391 668 | 1,059
.7 | 109 268 | 377
.8 | 16 47 | 63
-----------+----------------------+----------
Total | 15,885 8,877 | 24,762



*******************************************
End of the algorithm to estimate the pscore
*******************************************

DESCRIPTION OF VARIABLES:

des v136 v151 v152 hv217 v012 v109 v110 v112 v115 v119 sh39 v131 v137 v208
v218 v404 v416 husbwifelive v512 v627 v628 v629 v715 v705 v717 v718 v714
v731 v730 b9 s509 s510 s511a s511b s511d s512a s512b s513 sh29 sh32a sh32b
sh32c sh32d sh32e sh32f sh32x sh34 sh35 sh36 sh37 sh42 sh43 sh46 sh49 sh47a
sh47b sh47c sh47d sh47e sh47f sh47g sh47h sh47j sh47k sh47l sh47p sh47q
sh47r sh47s sh47t ownerTV hv211 v024

storage display value
variable name type format label variable label
-------------------------------------------------------------------------------
v136 byte %8.0g number of household members
v151 byte %8.0g v151 sex of household head
v152 byte %8.0g age of household head
hv217 byte %8.0g hv217 relationship structure
v012 byte %8.0g current age - respondent
v109 byte %8.0g v109 reads newspaper once a week
v110 byte %8.0g v110 watches tv every week
v112 byte %8.0g v112 listens to radio every week
v115 int %8.0g v115 time to get to water source
v119 byte %8.0g v119 has electricity
sh39 byte %8.0g sh39 religion of household head
v131 byte %8.0g v131 ethnicity (scheduled caste or
tribe)
v137 byte %8.0g number of children 5 and under
v208 byte %8.0g births in last five years
v218 byte %8.0g number of living children
v404 byte %8.0g v404 currently breastfeeding
v416 byte %8.0g v416 heard of oral rehydration
husbwifelive float %9.0g
v512 byte %8.0g years since first marriage
v627 byte %8.0g v627 ideal number of boys
v628 byte %8.0g v628 ideal number of girls
v629 byte %8.0g v629 ideal number of either sex
v715 byte %8.0g v715 husbands education-single yrs
v705 byte %8.0g v705 partner's occupation
v717 byte %8.0g v717 respondent's occupation
v718 byte %8.0g v718 current type of employment
v714 byte %8.0g v714 respondent currently working
v731 byte %8.0g v731 worked in last 12 months
v730 byte %8.0g v730 partners age
b9 byte %8.0g b9 who child lives with
s509 byte %8.0g s509 how much education should be
given to girls
s510 byte %8.0g s510 how much education should be
given to boys
s511a byte %8.0g s511a who decides what to cook
s511b byte %8.0g s511b who decides on obtaining
health
care
s511d byte %8.0g s511d who decides about respondent
staying with family
s512a byte %8.0g s512a permission needed to go to
market
s512b byte %8.0g s512b permission needed to visit
relatives or friends
s513 byte %8.0g s513 allowed to have money set
aside
sh29 byte %8.0g sh29 where do household members go
for treatment
sh32a byte %8.0g sh32a strain by cloth to purify
water
sh32b byte %8.0g sh32b use alum to purify water
sh32c byte %8.0g sh32c use water filter to purify
water
sh32d byte %8.0g sh32d boil water to purify
sh32e byte %8.0g sh32e use electronic purifier to
purify water
sh32f byte %8.0g sh32f use nothing to purify water
sh32x byte %8.0g sh32x use other method to purify
water
sh34 byte %8.0g sh34 main source of lighting
sh35 byte %8.0g sh35 number of rooms
sh36 byte %8.0g sh36 separate room used as a
kitchen
sh37 byte %8.0g sh37 main cooking fuel
sh42 byte %8.0g sh42 does this household own this
house or any other house?
sh43 byte %8.0g sh43 does this household own any
agricultural land?
sh46 byte %8.0g sh46 household owns livestock
sh49 byte %8.0g sh49 type of house
sh47a byte %8.0g sh47a mattress
sh47b byte %8.0g sh47b pressure cooker
sh47c byte %8.0g sh47c chair
sh47d byte %8.0g sh47d cot or bed
sh47e byte %8.0g sh47e table
sh47f byte %8.0g sh47f clock or watch
sh47g byte %8.0g sh47g fan
sh47h byte %8.0g sh47h bicycle
sh47j byte %8.0g sh47j sewing maching
sh47k byte %8.0g sh47k telephone
sh47l byte %8.0g sh47l refrigerator
sh47p byte %8.0g sh47p car
sh47q byte %8.0g sh47q water pump
sh47r byte %8.0g sh47r bullock cart
sh47s byte %8.0g sh47s thresher
sh47t byte %8.0g sh47t tractor
ownerTV float %9.0g
hv211 byte %8.0g hv211 has motorcycle
v024 byte %8.0g v024 region

.

_________________________________________________________________
Express yourself instantly with MSN Messenger! Download today it's FREE!
http://messenger.msn.click-url.com/go/onm00200471ave/direct/01/

*
* For searches and help try:
* http://www.stata.com/support/faqs/res/findit.html
* http://www.stata.com/support/statalist/faq
* http://www.ats.ucla.edu/stat/stata/

*
*   For searches and help try:
*   http://www.stata.com/support/faqs/res/findit.html
*   http://www.stata.com/support/statalist/faq
*   http://www.ats.ucla.edu/stat/stata/
_________________________________________________________________
Express yourself instantly with MSN Messenger! Download today it's FREE! http://messenger.msn.click-url.com/go/onm00200471ave/direct/01/

*
* For searches and help try:
* http://www.stata.com/support/faqs/res/findit.html
* http://www.stata.com/support/statalist/faq
* http://www.ats.ucla.edu/stat/stata/
_________________________________________________________________
Express yourself instantly with MSN Messenger! Download today it's FREE! http://messenger.msn.click-url.com/go/onm00200471ave/direct/01/

*
* For searches and help try:
* http://www.stata.com/support/faqs/res/findit.html
* http://www.stata.com/support/statalist/faq
* http://www.ats.ucla.edu/stat/stata/




© Copyright 1996–2024 StataCorp LLC   |   Terms of use   |   Privacy   |   Contact us   |   What's new   |   Site index