代写代考 HDAT9700 Statistical Modelling II

HDAT9700 Statistical Modelling II

# Overview

For your final assessment you are required to prepare a short analysis
report applying one of the statistical techniques covered in the course.
The report should be presented in IMRD
(Introduction/Methods/Results/Discussion) format and should be
approximately 1500 words.

## Topic options

You can base your report on one of the following three options:

**Causal analysis via matching** Exploring the causal effect of young
maternal age on perinatal outcomes

**Interrupted time series** Evaluating the effect of media coverage
following publication of new drug safety research on dispensing of oral
contraceptives using time series data

**Longitudinal multilevel modelling** Modelling children’s weight change
between 6 – 24 months

A dataset is supplied for each option. A more detailed description of
each option is provided at the end of this document.

## Report structure

Your report should be presented in IMRD format and be approximately 1500
words long. The word count does not include tables, figures, footnotes,
references, appendices, or the student declaration. You should include
up to five tables/figures in the main body of the report, with
supplementary tables/figures included in the appendix.

Below is in an example of how you might lay out your report:

1. Student declaration
2. Introduction (200-300 words approx.)
3. Methods (550-650 words approx.)
4. Results (350-450 words approx.)
5. Discussion (200-300 words approx.)
6. References (As appropriate)
7. Appendix (As appropriate)
8. Student declaration

### Introduction

*200-300 words approx.*

The introduction section should provide some brief background
information on the research area and motivate why the research question
is of scientific interest. Because you have been presented with a
research topic you do not need to make this section very detailed, you
should simply demonstrate that you have understood the relevance of the
research question, and make the connections between the research
question, the available data and the analytic approach.

### Methods

*550-650 words approx.*

In the methods section you should describe in detail the analyses that
were undertaken, and why. It is important that you demonstrate that you
understand the rationale for applying a particular statistical technique
and/or any alternative model specifications/model diagnostics. You
should demonstrate your understanding of the statistical methods with
appropriate references to the literature.

### Results

*350-450 words approx.*

In the results section you should begin with a descriptive overview of
the dataset you have used, including summaries of any key variables.
Follow this with a summary of your analysis results using appropriate
tables and figures, as well as describing your main results in text.
Appropriate model diagnostics and/or alternative model specifications
should be included but can be presented in the appendix if they are

### Discussion

*200-300 words approx.*

In the discussion you should summarise your main findings, interpret
your results in more detail, discuss implications, and acknowledge
strengths and limitations of the analysis, with an emphasis on the
statistical methods applied.

### References

Your report should include 5-10 appropriate references to the
literature, with a full list of references provided in the references
section. You can use your preferred referencing style. More information
on academic referencing can be found on the UNSW website:

### Appendix

The appendix should include any supplementary material that does not fit
in the body of the report, for example additional descriptive tables or
model diagnostics plots.

### Student declaration

All assessments should include a copy of the student declaration
regarding plagiarism, student academic misconduct and proper back-up of
your assessment. The text of the declaration is provided below. You
should copy this into your report and tick the boxes to indicate your
agreement with the statements.

> I declare that this assessment item is my own work, except where
> acknowledged, and has not been submitted for academic credit elsewhere
> or previously, or produced independently of this course (e.g. for a
> third party such as your place of employment) and acknowledge that the
> assessor of this item may, for the purpose of assessing this item: (i)
> Reproduce this assessment item and provide a copy to another member of
> the University; and/or (ii) Communicate a copy of this assessment item
> to a plagiarism checking service (which may then retain a copy of the
> assessment item on its database for the purpose of future plagiarism
> checking).
> – \[ \] I understand and agree I certify that I have read and
> understood the University Rules in respect of Student Academic
> Misconduct.
> – \[ \] I understand and agree I have a backup copy of the
> assessment.
> – \[ \] I understand and agree

## Accessing and submitting your report

The data for each assessment option, and any additional hints or
instructions, will be made available in a GitHub repository. You can
clone this repository through a hyperlink provided by your course
convenor. To submit your assessment files, add and commit them to the
repo, and then push to GitHub. You are welcome to explore all three
assessment options before deciding which option to choose. You may
submit your assessment as either:

1. A Microsoft Word document with the corresponding R code provided
separately
2. an R Markdown file, interweaving the text of your report,
statistical code and results.

## Statistical code

You should carry out all data preparation and statistical analyses using
R. Include all R code you use to prepare and analyse the data as part of
your submission. Following best practice in reproducibility, it should
be straightforward for the marker to reproduce your entire analysis
using only the raw data and the code you submit. Your code should be
clearly commented, to assist in this reproducibility, and to demonstrate
your understanding of the analyses being performed.

## Marking rubric

Your assessment will be marked according to the marking rubric on the
following page. Four criteria are assessed: • Conceptual understanding
(30%) • Application of statistical method (30%) • Interpretation (30%) •
Presentation (10%) 6. Plagiarism

While you are welcome to discuss the assessment with your classmates,
the final written report and statistical code must be your own work.
References to other scholarly work should be properly cited. **Direct
plagiarism will not be accepted.** Past incidences of plagiarism have
resulted in students retaking the assessment. You can read academic
integrity at UNSW and examples of plagiarism here:

## Submission deadline and late submissions

Assessment 2 is due by **9am on Monday 15 August 2022**. A penalty of 5%
per day will apply to late submissions. To allow marking to take place
in time for the university grade submission deadlines, late assessments
cannot be accepted after Thursday 18th August (barring special
circumstances). Please notify the course convenors as soon as possible
in the case of extenuating circumstances that prevent submission by this

————————————————————————

# Option 1: Causal inference via matching

## Overview

Young maternal age is associated with adverse pregnancy outcomes,
including low birthweight and preterm birth. As a result, in some
contexts “teenage pregnancy” has been framed as a public health problem.
However, many academics and health professionals have argued that young
maternal age doesn’t cause adverse outcomes, but rather socio-economic
disadvantage is the primary factor underlying different perinatal
outcomes experienced by younger and older mothers. (cf. Lawlor and Shaw
(2002) “Too much too young? Teenage pregnancy is not a public health
problem”, International Journal of Epidemiology (31) p552-554, and the
subsequent counterpoint articles, for more on this debate).

In this assignment you will use data on maternal and family demographics
and child birth outcomes, to estimate the causal effect of young
maternal age on the risk of preterm birth. Use the MatchIt package in R
to identify a matched “treatment” and control group, who are as similar
as possible in terms of appropriate pre-treatment variables. A DAG
should be used to guide variable selection and highlight limitations of
the available data.

## Research question

> What is the causal effect of young maternal age on the risk of preterm

The data for this assessment are drawn from the National Children’s
Study Archive. The National Children’s Study (NCS) collected birth and
early childhood data on more than 5,000 children and their families in
the USA from 2009-2014. The NCS teaching database contains base
child-level data, as well as several topic-specific modules.

You are provided with a dataset named **ncs_child_mom_matching.csv**.
This file contains information on 3,848 singleton births including
mother and child demographics, birth outcomes and area-level statistics.
The data are drawn from the NCS base Child dataset, and the Mother’s
Pregnancy Health module. Some data preparation has already taken place,
in particular, the aforementioned Child and Mother datasets have been
merged, and variables with high proportions of missing data have been

The following resources will be useful to help understand the data: \*
Information on the National Children’s Study Archive \* NCS study
description and guide \* Codebook for the base Child data \* Codebook
for the Mother’s Pregnancy Health module Hints \* Young maternal age is
usually defined as maternal age less than or equal to 19 years. \*
Preterm birth is usually defined as gestational age less than or equal
to 37 weeks. \* Use the MatchIt package in R to explore different
matching approaches. \* Assess balance between the treated and matched
control groups using numeric and graphical summaries. \* Not all
variables are necessarily appropriate for matching and not all variables
you would like to match on will be available. Use DAGs to guide variable
selection and highlight limitations. \* Remember to apply the matching
weights when analysing matched data.

————————————————————————

# Option 2: Interrupted time series analysis

## Overview

Oral contraceptives (“birth control pills”) are taken by women to
prevent pregnancy and to treat other conditions such as endometriosis or
polycystic ovary syndrome. There are two main types of oral
contraceptive pills, the most common being the combined pill, which
contains a combination of the hormones oestrogen and progestogen in
various formulations and doses. The second type is the progestogen-only
pill (also called the “mini pill”) that contains only progestogen.

Although they are effective at reducing the occurrence of unwanted
pregnancy, the combined pill increases the risk of blood clot formation,
such as deep vein thrombosis, pulmonary embolism and stroke. A paper
published in the BMJ on 26 May 2015 was the first to quantify the risk
associated with different oestrogen/progestogen combinations and
received substantial media attention worldwide (Vinogradova et al. Use
of combined oral contraceptives and risk of venous thromboembolism:
nested case-control studies using the QResearch and CPRD databases. BMJ
2015;350:h2135). Note that there is no increased risk of blood clots in
women using the progestogen-only pill.

In this assessment you will use time series data to explore the impact
of the media attention surrounding the publication of this study on the
PBS-subsidised dispensing of combined and progestogen-only oral
contraceptives. In Australia, combined contraceptive pills containing
levonorgestrel/ethinylestradiol, norethisterone/ethinylestradiol and
norethisterone/mestranol are subsidised through the PBS. Other combined
pill formulations are not subsidised, such as those containing
drosperinone, cyproterone and desogestrel and are not captured in the
data. Several progestogen-only mini pills are also subsidised, including
levonorgestrel and norethisterone.

## Research question

> What was the impact of the media attention around the increased risk
> of thrombosis associated with oral contraceptive use on dispensing of
> combined oral contraceptives?

Data are provided in the file contraceptives.csv. This file includes
monthly counts (per 10,000 women of reproductive age) of PBS-subsidised
dispensings of combined oral contraceptives (“combined”) and
progestogen-only contraceptives (“mini”) between Jan 2013 and Dec 2016.
The peak of the media attention was in the last week of May 2015. Hints
\* Use appropriate visualisations to explore the raw data. \* Explore
seasonal effects, autocorrelation, and delayed impacts with statistics
and tests. \* Create appropriate variables to model changes after the
intervention. \* Use an appropriate model to explore whether there was a
change in combined oral contraceptive dispensing following the media
attention. \* Plot the predicted counterfactual to visualise any change
in dispensing following the media attention. \* Use the “control” series
(dispensing of progestogen-only oral contraceptives) and comment on the
findings in relation to those from the combined pill series.

————————————————————————

# Option 3: Longitudinal multilevel modelling

## Overview

Children’s growth is frequently monitored during their infancy and
understanding growth trajectories is important for identification of
chronic disorders or other health issues, reassuring parents and
providing information on the health of a nation’s children.

In this assignment you will use data including repeated measurements of
children’s weight to model children’s weight change over time. The data
also include information on other infant, environmental and maternal
factors, and you will also explore which of them could further improve
the model for children’s weight gain. You should use the nlme package in
R for carrying out your longitudinal multilevel modelling.

## Research question

> What is the optimal longitudinal multilevel model for modelling
> children’s weight change?

You are provided with a dataset named **ncs_child_repeated.csv**. This
file contains information on child, mother and household demographics
for 4,531 kids, with 1-4 weight observations for each child (12,654
observations overall). The data are drawn from the NCS base Child
dataset, and the Child health module, including children’s weight
measurements at 6, 12, 18 and 24 months after their birth. Some data
preparation has already taken place, in particular, the above files have
been merged, only singleton births and the oldest siblings have been
included (i.e. only one child per mother), and variables with high
proportions of missing data have been removed.

#### Hints

– The outcome weight (`visit_wt`) can be modelled as a function of
continuous age (`child_age`) or ordinal visit timing (`visit`).
– Conduct appropriate EDA to understand the relationship between
weight and age.
– Use the `nlme` package for carrying out your multilevel modelling,
starting with simple models and progressively exploring more complex
structures.
– There is only one child per mother, and thus you do not need to
include mother as an additional level in your multilevel modelling.
– Explore alternative functions of time to improve model fit.
– Explore other covariates that help to explain variation in
children’s weight.
– Some child health variables (e.g. `gastro`, `diarrhoea`,
`ear_infection`) were not measured on every visit occasion, so if
these are included you will need to think about how to handle that.

————————————————————————

# Marking rubric

HDAT9700 Statistical Modelling II, Assessment 2 Marking Rubric

Related Posts