Questions to PI

[kaedonkers]

(mariyam-khan) 1st question: What is definition and measure of social inequality in health? 2nd question: Which methods in deep learning to utilize? 3rd question : What is counted as European country? 4th question: What is the cost related to the project/budget?

[kaedonkers]

Data Science Project Lifecycle - questions to consider: https://alan-turing-institute.github.io/rds-course/modules/m1/1.2-DataScienceProjectLifecycle.html

Question 1: What is the broad problem we are trying to solve? Question 2: What is the specific research question? How does it translate to a data science problem? Question 3: Is data available and appropriate? Question 4: What are the stakeholders’ expectations? Question 5: How does the output product look like and how is it going to be used? Question 6: What is the state of the art? Question 7: What is in-scope and out-of-scope? Question 8: What is the expected impact? Question 9: How do we measure the success of the project? Question 10: How do we monitor and explain output? Question 11: What computational resources are available? Question 12: How is the product going to be deployed and maintained? Question 13: What are the timelines and milestones of the project? Question 14: Ethical considerations

[aleonedecastris]

What is the added value of this study relative to existing literature?

[Ranapop]

Q & A

Data and countries considered

Q: What is the geographical scope A: We want to do all the countries included in this study. We want to look at all the countries surveyed in the 2016 survey. The link I first sent is to the 2016 data, but that is not actually open access.

Q: Can we have access to the data? A: No

Q: Can we use the 2007 and 2011? A: Yes

Q: Which countries? A: All the countries

Q: Do you want to see comparisons between the 2 waves? A: Yes.

Q: What do we do with the countries that we only have 1 wave on? A: Include them but exclude comparisons. We can also look in only one wave or only at the common countries.

Research question

Q: Do we have an actual research question A: Yes, why do these inequalities exist? What's driving them? You should be looking for correlations.

Q: Are there any variables we should focus on? A: Social material, occupation and psychological factors that might affect inequalities.

Q: Do we have restrictions/requirements on what to compare? (eg. waves, regions) or is it up to us? A: Up to us.

Health inequality

Q: How do we define health inequality? Can we use the standard methods from the literature? A: Yes, I am happy with that

Data aggregation

Q: The data is on individual countries, how can we look at the whole of Europe? A: There is a weighting scheme.

Cost

Q: What does the digital component include? A: Staffing (hours per week), computational time

Expected product

Q: What is the result of the work? A: Some type of report (published in a journal), maybe output some code that we could produce (like an interactive way of looking at the data). I would like to influence the government with this so I would like a couple of good papers.

Q: Should we focus more on papers or on artefacts? (like the interactive app) A: The app could be part of a follow-up project.

Q: What governments do you want to present to? UK or EU? A: All governments.

Q: Any other stakeholders? A: just governments

Q: Is anything out of scope? A: What we really want to focus on is Europe (also in terms of data). If there was a way of looking at some averages of some places within Europe that would be good. It would be interesting to do a comparison across time but if there is data missing we can just look at the existing countries.We can look at correlations but the postdoc can look at that.

Q: What is the role of the Postdoc? A: You can work with the postdoc, but he will mainly be doing my work

Funding

Q: Who is funding this? A: I will be writing a grant -- public.

Methods

Q: Any methods we should focus on? A: You should find the best methodologies for this and you can come back to me with that. I think DL is the way forward but I am open to other methods.

Demographics

Q: Will the study impact any demographics we should consider (ethically)? A: As the data is public this should be ok, but we should make sure we are not making? any wrong assumptions.

Plan

Q: Is the first step looking into methodologies? A: The first thing to do is think about methodologies and set up a plan. For eg, look at this particular regions of Europe, look at this particular measures of inequalities and then narrow down on some variables.

Literature review

Q: Should we focus on something the existing literature hasn't covered yet? A: We can look at similar things, but look at the whole of Europe and not only at specific countries.

Short time plan

Q: What is the main goal in the next 2 weeks? A: The main thing I need is the costing 1) How much time 2) How many people

Q: Should we come up with a budget plan for staffing, computational resources, office space etc? A: Yes, hopefully you can get that back to me in 2 weeks.

Type of health

Q: Do we consider physical or mental health? A: Both

Publications

Q: How do we measure publication impact? Do you have a journal in mind? A: Haven't thought of that, but it should be a high-impact in an open access journal.

Computational capabilities

Q: What do you have access to? A: I will apply in my funding to get access to storage, machines. That's something I need to apply for.

Q: So this is not restricted yet? So should we provide you with the information of how much storage and compute power we need? A: We can look together into that, it probably depends on what methods we want to use (eg DL). I would be open to buy a website if we wanted to create that interactive app.

Goal

Q: Is our goal to look into policies that governments could use? A: No, your goal is more to look into the data. My postdoc will look into policies. What I want from you guys is to find those correlations.

Stakeholders

Q: So we do not have any particular stakeholders? A: No, the idea is to eventually get the governments involved but for the moment it's just research.

Expected model

A: We want a predictive model. Say we have some data from another country (but not the health data) and we would some health inequality index.

Q: So we would initially look at correlations between vars and then turn it into a supervised learning setting? A; Yes

Measures from data

A: The data records self-reported health and that is one of the measures we could be using. We should keep in mind that this is self-reported.

Correcting for self reporting

Q: How do we deal with the measures being self-reported? If in some countries for eg. people "exaggerate" more? A: We can look at objective measures to correct for that.

Measures of success

Q: Journal? A: High impact journal

Q: No of citations after a year? A: More than the avg of the related work(?)