代做Observational Data、代写R程序、R编程代做、代写Modelling Prediction

日期：2018-11-25 09:16

Modelling Prediction and Causality with Observational Data University of Leeds

Arnold KF, Gilthorpe MS (Leeds) Page | 1

ASSIGNMENT 1

The government has commissioned a study into the performance of A&E departments across

England. Of top priority is creating a model (1) that ‘red flags’ A&E departments – i.e., identifies

those with an unusually high average weekly number of A&E attendances per capita. Such a model

can be used to predict potential ‘red flag’ A&E departments, which will be put on a ‘watch list’ for

special measures and will warrant further investigation into why unusually elevated attendance

rates occur. The long-term aim is to curb such excessive use of A&E facilities by finding ways to

deal with as many health issues preventatively as possible via general practitioners or pharmacies.

The government also wants a model (2) that predicts the average weekly number of A&E

attendances in each department to help target resources and aid resource planning across

departments. This model may utilise as an outcome either A&E attendances or A&E attendances

per capita; as a researcher, you must consider which is the more appropriate to meet the

government’s needs and justify which you have selected.

To create the two models, you have been provided the dataset ‘AEdata 2018.csv’, which comprises

the following variables:

AEn: average weekly number of A&E attendances

AEpc: average weekly number of A&E attendances per capita (average weekly number of A&E

attendances / population)

Flag: ‘red flag’ department (0 = no, 1 = yes)

Area.Type: area type (0 = rural, 1 = urban)

Area.Size: area size (km2

)

Pop: population (in thousands)

Pop.dens: population density (population / area size)

GPn: number of GP practices

PHARMn: number of pharmacies

PHARMpc: number of pharmacies per capita (number of pharmacies / population)

You must write a formal report (1000 words maximum) that summarises your findings for both

models. Your report should explain which covariates you consider for your models to predict each

of your outcomes, and why you consider these. Your report must also include basic summary

statistics of the data you have been provided, in addition to a more detailed explanation of the two

models you generate. You must justify: (a) which fit criteria you use to select your models; (b)

which continuous outcome you have chosen to model (i.e. A&E attendances or A&E attendances

per capita); and (c) all other decisions you have taken as a researcher to arrive at the final models

you report. You should discuss the strengths and weaknesses of what you have done and propose

potential improvements for future modelling.

You should pay special attention to presentational issues; how you present your findings is (nearly)

as important as the findings themselves. You might explore the public domain for similar

documents. You must provide clarity of language. The report may have tables and figures in the

main text, but these must not exceed two of each (and will not count towards the word limit).

Attach your annotated R code as an appendix, and include (Harvard style) citations to justify

decisions you make or to place your work in the wider context in a bibliography (these will not

count towards the word limit).

Marks out of 50 will be awarded as per the following criteria: (a) clarity in the development and

justification of your predictive models; (b) detailed exploration and justification of the criteria used

for assessment of model fit; (c) well-structured presentation and clear language that explains and

discusses all you have done, including appropriate use of citations and appendices; (d) discussion

of the strengths, limitations, and future recommendations.