Assignment 1, PSA
Problem 1: Multiple Choice Questions (2 points for each problem)
1. Counterfactual is synonymous with ________.
a. potential outcome
b. counterfactual framework
c. unobrved heterogeneity
d. both a and b
2. The treatment effect model shares same features as the propensity score matching in
the following way(s) ________.
a. both estimate propensity scores
b. both create a matched sample
c. both cau reduction of sample size
d. both a and c
3. The strongly ignorable treatment assignment assumption is synonymous with
________.
a. no omission of important predictors of sample lection
b. independence between an independent variable and the error term of a regression
model
c. no correlation between treatment assignment and a covariate of outcome
d. both a and b
e. all a, b, and c
明太鱼的做法4. The challenge(s) of running instrumental variable approach is (are) the difficulty (or
difficulties) to________.
a. know potential outcome for all study participants
b. know the caus of sample lection
c. to find a uful instrumental variable
d. all a, b and c
e. both b and c
5. The Heckman sample lection model differs from the treatment effect model in the
following way(s) ________.
a. estimation methods (i.e., two-step estimator versus maximum likelihood)
b. whether the regression equation includes the binary treatment variable
蜀江烤鱼c. whether the outcome variable is known for all participants
d. both b and c
e. all a, b, and c
6. The SUTVA implies that ________.
a. to test the effectiveness of aspirin the evaluator should give aspirin to all study
participants
b. to test effectiveness of aspirin the evaluator should u exactly the same type of
aspirin in the study
c. recent graduates from a job training program should not affect the average income
in the job market
d. all a, b and c
e. both b and c
7. When Rubin (2008) says “design trumps analysis”, he means that ________.
a. a randomized experiment should always be designed carefully
b. an obrvational study should model randomization and make efforts to correct for
nonrandomized lections
c. analysis is less important than design when conducting an obrvational study
d. all a, b and c
e. both a and b
8. The conquence(s) of having an endogeneity problem in running regression is (are)
________.
a. the estimated R2 is misleading
b. the significance tests are misleading
dm海报
c. treatment effect is exaggerated
d. all a, b and c
e. both b and c
9. The randomized experiment (RE) is superior to an obrvational study (OS) in the
following way(s): ______________
a. RE is less expensive
b. OS fails to control for unmeasured heterogeneity
拓跋玉>土味情话
c. RE do not require restrictive assumptions
d. all a, b and c
e. both b and c
10. Which of the following statement(s) is (are) fal? ______________
a. average treatment effect equals to intent-to-treat effect
b. efficacy effect is the same as the treatment effect for the treated
c. in a randomized experiment the average treatment effect for the treated is the
same as the average treatment effect for the control
d. all a, b and c
e. both b and c
Problem 2 (30 points)
In the data file “chpt4_1.dta” of the PSA Website, the variable “bc3_tpt” is a total score of CBCL. Like the externalizing and internalizing scores, a high value on the total score indicates a greater extent of behavioral problems.
U this variable as an outcome, and exactly the same t of covariates affecting the lection and regression equations as tho for the externalizing model (i.e. the model prented in Section 4.5.1 of Guo & Frar, 2009, pp108-113), run a treatment effect
model using maximum likelihood estimator. Attach (or copy/paste) the output to your completed assignment. Answer the following questions.
2.1 What is your model estimated treatment effect? Interpret your finding.
2.2 What is your estimated correlation between the error of the lection equation and the
error of the regression equation? Perform a significance test of the null hypothesis stating that the correlation is zero. What do your test inform you?
2.3 Interpret the meaning of model chi-square. What does the test of the significance of
model chi-square inform you?
2.4 From your estimated model, what do you know about important covariates that affect
the sample lection? What does sample lection mean under the current context? 2.5 What variables are important in predicting participants’ total score?
2.6 Run the same model using a two-step estimator.
2.7 Compare results between maximum likelihood method and the two-step method.
Examine each covariate, and then report whether or not the two estimators give you the same findings. Note that by same findings, I mean the same sign of regression and lection coefficient and same conclusion of significance test (two-tailed) at a .05 level.
馄饨煮几分钟
Problem 3 (20 points)
You are provided with 50 imputed files in which missing values of independent variables are all imputed. You can download the data (i.e., the zip file “Data_4.5.3.zip”) from the PSA Website and unzip it. Run a treatment effect model bad on all 50 files with the following specifications:
Regression equation:
通用学术英语
icsaggch = β0 + β1 ageyc + β2 fmale + β3 blck + β4 whit + β5 hisp + β6 pcedu + β7 ipovl + β8 pcemft + β9 fthr + ε
intbl = β0 + β1 ageyc +β2 fmale + β3 blck + β4whit + β5 hisp +β6 pcedu +β7 ipovl + β8 pcemft + β9 fthr + β10 dicsagg2 + β11 dicsint2 + u
照相机发明时间
The same model should also control for clustering effect. That is, u schbl as a clustering variable (i.e., a variable denoting school ID—study participants are nested within schools) to obtain a robust estimate of standard errors to correct for clustering effect.
Report:
3.1 The estimated treatment effect bad on all 50 imputed files in terms of the magnitude
of the effect and whether or not it’s statistically significant.
3.2 Which variables might be important in affecting sample lection? Be clear which
significance level you u in the evaluation.
3.3 Which variables might be important in affecting outcome? Be clear which
significance level you u in the evaluation.
3.4 What is the estimated correlation between εand u?
Problem 4 (30 points)
Write an abstract (< 400 words) to describe a design of obrvational study. You may choo a rearch problem from the substantive area of your interest. U Rubin’s (2008) general advice to show important components of your design.
