ECON5570 HEALTH ANALYTICS ASSESSMENT #2

r>

This assessment will be on the topic of immunization. It is an individual

assessment comprising 45% of your final grade. The assessment has three parts

which requires you to write an essay and analyse and interpret two datasets using

different methods of statistical analysis. It will cover materials on the second and

third part of the course relating to regression analysis and causal inference.

Specifically, the assessment will draw upon country level immunization data

from the World Bank and data from a randomised controlled evaluation of

immunization campaigns in rural India adopted in the study by Abhijit Banerjee

and colleagues, entitled “Improving immunisation coverage in rural India:

clustered randomised controlled evaluation of immunisation campaigns with and

without incentives” published in BMJ (2010). It is recommended that you read

this article carefully before commencing the assessment. The datasets are

provided in the assessment materials in the LMS (they are in Stata format so if

you’d like to import them into R follow these instructions).

Please give clear concise answers (the least amount of words it takes to make

your meaning clear). The point of this assessment is to think thoroughly about

the models and issues discussed in the lectures and workshops. Where

appropriate your answers should include the relevant analysis output from the

statistical software of your choosing. For example, you should include the

regression results output table rather than just reporting the coefficient value in

the text.

Feel free to work in groups. However, you must hand in individual assessments,

explaining the answers in your own words. Remember the university’s policy on

plagiarism applies, this must be your own work.

Part 1. Immunization Essay

Provide an overview of immunization around the world and its role in improving survival in

children. What is the recommended package of basic immunization for children and what

diseases does it provide protection from? What are the population health benefits of

immunization? What are the consequences for children and pregnant women who do not

receive immunization? What do we know about the level of inequality of immunization rates

around the world and what are some of the reasons for varying coverage rates between

developed versus developing countries? Your essay should not exceed 500 words and should

draw upon references beyond Banerjee et al. (2010).

[Total 5 marks]

Part 2. World Bank Immunization Data

To complete this part you will need to use the World Bank immunization dataset which

includes data on the immunization rate against measles in 172 countries, among children of

age 12 to 23 months old, from 1998 till 2017. Key variables in the dataset include

immunization rate, child survival rate, GDP per capita and population size.

(i) Restrict the data to the most recent wave (2017). What are the mean and standard

deviation of GDP per capita, immunization rate and child survival across countries?

What do they tell us? [2 marks]

(ii) Now present the mean and standard deviation for the same three variables separately

for the poorest 50 countries and the richest 50 countries. What does this tell us about

the relationship between income and immunization and income and child survival? [2

marks]

(iii) Graph a scatterplot of the joint distribution of the rate of immunization and GDP per

capita. Do this first in levels of GDP per capita and then in natural logarithm form.

Provide a description of the relationship between GDP and the rate of immunization

in the two graphs. How do they differ and why? [2 marks]

(iv) Regress the rate of immunization on GDP per capita. Based upon your graphical

analysis in (iii), provide an explanation of your preferred specification of the

relationship between GDP per capita and immunization rate. Graph a scatter plot of

the data as well as the regression line. Interpret the coefficient on GDP per capita.

What is the t-statistic for this coefficient and what does it tell you? Interpret the 95%

confidence interval. [2 marks]

(v) Explain why this may be considered a naïve estimate of the relationship between

immunization and GDP per capita? What are the possible sources of bias? Provide an

example of omitted variable bias and how will it influence the estimated coefficient

on GDP per capita [2 marks]

(vi) Let’s now look at the relationship between child survival and the rate of immunisation

across countries. Regress the child survival rate on the rate of immunisation across

countries. Report the coefficient on the immunisation rate and its standard error; what

do they tell you? Is the sign of the coefficient what you expected? Explain briefly. [2

marks]

(vii) Based upon your analysis above does it make sense to control for GDP per capita in

the model? Why or why not? Report the regression results when controlling for GDP

per capita. How has the coefficient on immunisation coverage changed? Why is this

so? What happens when you control further for the log of population size? Explain. [2

marks]

(viii) Return now to the original panel dataset, years 1998 – 2017 in 172 countries. Using

OLS, regress child survival on immunisation coverage, GDP per capita, population

size and year using the appropriate specification of the functional form of the

variables. How has the coefficient changed on immunisation compared to the same

model estimated using a single cross-section above and why? [2 marks]

(ix) Now estimate the same specification with an additional country category variable

(country fixed effect). What does the coefficient of interest on immunisation now

measure? How does it differ to that generated in the above regression? [2 marks]

(x) Based upon the above regressions, what is your preferred estimate on the effect of

immunisation rate coverage on child survival and why? Is it a causal estimate and

why or why not? [2 marks]

[Total 20 marks]

Part 3. Improving Immunization Coverage in Rural India

This part focuses on the paper by Abhijit Banerjee and colleagues (2010), “Improving

immunisation coverage in rural India: clustered randomised controlled evaluation of

immunisation campaigns with and without incentives.”

(i) Briefly explain the aim and contribution of the paper. What question does it ask?

How does it go about answering the question? Briefly explain the study design and

methodology. What is unusual/exciting about the method relative to the above

methods we have applied? Your response should not exceed 500 words. [5 marks]

(ii) Under random assignment, the expected value of the outcome of interest and

variables which influence the outcome for the treatment and control group are

assumed to be equal in the absence of the intervention. Using the baseline dataset

show that the baseline characteristics of the control and treatment groups are

comparable and not statistically different from zero. [5 marks]

(iii) Using the endline dataset, estimate the average treatment effects of the intervention

on immunisation outcomes. First, show the simple difference in mean outcomes

between treatment and control groups. Second, regress immunisation outcomes on

the treatment indicator variable/s. Third, regress immunisation outcomes on the

treatment indicator variable and control variables including age of the child and

block of residence. In each case interpret the estimated treatment effect. Does the

coefficient or standard error change when controlling for other variables in the

model? Is this in line with your expectations? Why is this so? [5 marks]

(iv) Summarise the main findings of your analysis above and the implications for policy

in consideration of the issues that you raised in your immunization essay in Part 1.

In addition, discuss some limitations of the study and possible threats to the

robustness of the results. Are there any possible threats to internal and external

validity that you can think of? Your response should not exceed 500 words. [5

marks]

[Total 20 marks]

欢迎咨询51作业君

Email:51zuoyejun

@gmail.com

添加客服微信: **ITCSdaixie**