this an old idea from urban planning & urban economics, sometimes also referred to as economies of scale

interesting

good source - you may find other works by this author that investigate urbanization as a complex system and cities being described in terms of their fractalization and power law distributions

good

good

good

significance

good

These are superb annotations. Your theme and focus has been clearly established. Topically boundaries also defined. A few thoughts.

(1) Prioritize the methods from those that you think present the most promise to the least. Your review has captured some examples that border on both machine learning as well as spatial statistics. This is really impressive. You have also accomplished this very definitely within the context of a clearly articulated human development problem. Just keep looking for new, state of the art methods that present the most promise. Are you familiar with AI Nigeria? GRID3 Nigeria? There is really interesting stuff coming out of those groups in conjunction with Southhampton and Columbia.

(2) Try to narrow your focus a bit. Are you focusing on the intersection of poverty reduction and food security? education? gender equality? Formally articulating your central research question promote this continued narrowing of your focus.

Just keep going!

super!

excellent

excellent - identifying interactions is really important. Are you taking stats?

development is not the only outcome that may emerge from Nigeria's highly complex and adaptive, social and economic system

good

Great source. Focusing upon gender inequalities is arguably the most significant first step to take in order to reduce poverty.

excellent

good

good source

excellent

excellent

good

good

interesting

good

significance - good

harms & significance - good

This is excellent work. Some thoughts.

(1) You have identified a wide array of potentially viable and effective paths forward. You might want to prioritize the many methods you have identified.

(2) Can you further describe the methods in detail? How is the CDR data being used? How do gravity, radiation, and impedance models functions? What are voronoi polygons? What is state of the art for using CDR data to describe, analyze, model and then predict human movement during an infectious disease outbreak? Just keep pushing forward with your analysis in this area. Also, have a look at flowkit. There's a GitHub repository that will be useful.

(3) I think you could continue to focus on describing the process itself. Why is it important to model human behavior using cell phone data? How does understanding human movement help us to better address an outbreak? How could this information be used in a time of crisis to expedite solving the problem? You've have captured multiple very significant ideas that go straight to inherency, but now organize and structure them in order to describe the process to better inform response.

Just keep going!

super

good

excellent

excellent

excellent

great sources

good

good

good

excellent source

good, are you able to further describe the method?

good

good

good source

OK good. Some thoughts.

(1) Are you able to identify some machine learning or other data science methods that IF used in the context of your research and location would present promise towards better understanding your problem? Perhaps a random forest approach from world pop could be used to spatially locate households and persons? Perhaps a hierarchical bayesian model (also from world pop) could be used to describe demographic and economic attributes of persons and households? How would such "what if" applications add value to your investigation?

(2) Can you further articulate your problem statement? Which aspect of development will you focus upon with your continued investigation? Have you identified a gap in the literature? How about a tentative research question?

Just keep going!

interesting

good

good

good

how interesting

Super. Some thoughts.

(1) You could probably go two different directions with this investigation. I think a lot of the current research regarding malaria and vector borne diseases revolve around using CDR data with gravity models, radiation models and impedance models. You may also find an agent-based model of transportation systems and human movement that incorporates predictions of malaria prevalence. It could also be possible to focus on the natural environmental raster data (precipitation, water, climate factors) as predictors in a machine learning method, which is more of the direction you have taken with these annotations. While spread of the disease can be predicted by modeling transport behavior, at its origin, most countries that have eradicated malaria and similar types of vector borne diseases have done so through public works efforts that seek out to spray insecticides that kill mosquito larvae in standing water as breeding areas (is this the case in Cambodia?). If it is possible to more accurately (spatially) and then effectively identify these source locations, potentially there could be great progress towards eradicating the disease.

(2) I have also seen a few studies that seek to model the mosquito population itself. I think the limitation of that approach has been largely computational due to the sheer numbers, the scale of the insect and the environmental and natural factors associated. Maybe that has recently changed.

(3) I still wonder if climate change is increasing malaria disease prevalence. Are the mosquitoes and parasites manifesting in locations across Africa where they previously hadn't been observed? Are they more prevalent in China? In other locations throughout the globe?

(4) Are there any other machine learning or data science methods that IF used in the context of your research and location would present promise towards better understanding your problem? Perhaps a random forest approach from world pop could be used to spatially locate households and persons? Perhaps a hierarchical bayesian model (also from world pop) could be used to describe demographic and economic attributes of persons and households? What about an agent-based model? How would such "what if" applications add value to your investigation?

Just keep going!

did they use a model? perhaps CDR data?

good source

good platform from Luke Anselin

interesting

good

I think this is the same guy who is famous for using STATA, which for a long time was the biggest competitor to R as statistical software (I used STATA for a long time before switching to R)

good data source

good

interesting, goes to inherency

good

excellent source and annotation, a bit dated but still useful and effective, especially as a direction marker

excellent

interesting

good

good to consider both

excellent

good

Excellent work. Some thoughts:

(1) Narrow your focus in terms of the problem you are seeking to address. Investing a bit more time with regard to your problem statement will help you to narrow your focus and further articulate your research question.

(2) Kriging as a data science method is a good path forward. It will be interesting to see what you come up with in terms of how this spatial statistics methodology is reinforced with machine learning. Use of point process models (Baddeley) is another area of active research that might be find usefulness in the context of your work. Spatial random forest is a kind of next generation spatial statistic methodology that synthesizes machine learning.

(3) Are there any other machine learning or data science methods that IF used in the context of your research and location would present promise towards better understanding your problem? Perhaps a random forest approach from world pop could be used to spatially locate households and persons? Perhaps a hierarchical bayesian model (also from world pop) could be used to describe demographic and economic attributes of persons and households? What about an agent-based model? How would such "what if" applications add value to your investigation?

Just keep going!

very good, do you think a machine learning approach might present additional value? perhaps random forest, or spatial random forest?

good interpretation

ok

good spatial description

good

good

Excellent work. Some thoughts:

(1) Narrow your focus in terms of the problem you are seeking to address. Degradation of water quality will likely involve questions that address either non-point or point sources of pollution. Which ones are the most significant offenders? You could also narrow your focus in terms of water use. Are you interested in potable water? Urban water use? Agricultural water use? Investing a bit more time with regard to your problem statement will help you to narrow your focus and further articulate your research question.

(2) Kriging as a data science method is a good path forward. It will be interesting to see what you come up with in terms of how this spatial statistics methodology is reinforced with machine learning. Use of point process models Baddeley) is another area of active research that might be find usefulness in the context of your work.

(3) A superb use of a webpage to produce your annotations. Please confirm the links on your index page are functional.

Excellent thematic focus. Clearly defined topic. A few thoughts.

(1) Are you able to identify some machine learning or other data science methods that IF used in the context of your research and location would present promise towards better understanding your problem? Perhaps a random forest approach from world pop could be used to spatially locate households and persons? Perhaps a hierarchical bayesian model (also from world pop) could be used to describe demographic and economic attributes of persons and households? Perhaps looking at Utazi, Alegana and some other authors on WorldPop and flow minder might present some options. You do have the Alegana paper from 2015 here, which is good. I think hierarchical bayesian model present the most promise for your investigation. How would such "what if" applications add value to your investigation?

I'm going to send you a previously students literature review on this topic to support your work.

(2) Are you able to identify a gap in the literature? Are you able to formulate a research question that aligns with that gap? Have you classified that research question as a type of "puzzle" that you are attempting to solve?

Just keep going?

good, are you able to further describe their model?

good source

excellent

Excellent problem statement - one comment, GIS typically refers to software that ultimately enables users to advance spatial description, analysis and in some cases also models and then prediction (forecasts). It is also typically very effective at communicating spatial data. Spatial data is by nature very complex, thus for a long time, GIS has simplified the complex nature inherent to this type of analysis. At its core, you can think of GIS techniques translating as spatial description, analysis, modeling and prediction -- or perhaps more simply put, spatial statistics. Adding the human or environmental dimension, simply further defines the domain of this spatial inquiry by adding a geo- prefix to the spatial, thus geospatial statistics.

wow, it really is staggeringly high

good

OK good. Two interesting possible directions. Some thoughts.

(1) Are you able to further describe one or more of the machine learning or other data science methods that IF used in the context of your research and location would present promise towards better understanding your problem? Perhaps a random forest approach from world pop could be used to spatially locate households, persons and accessibility to primary, secondary or higher education services? Perhaps a hierarchical bayesian model (you mentioned two different ones in your annotation) could be used to describe demographic and economic attributes of persons and households? How would such "what if" applications add value to your investigation?

(2) Are you able to identify a gap in the literature? Are you able to formulate a research question that aligns with that gap? Have you classified that research question as a type of "puzzle" that you are attempting to solve?

Just keep going!

good

interesting

excellent

interesting, I hadn't seen this study before -- a bit dated, but good for your thematic / conceptual focus

OK good. Nice thematic focus and problem statement. Likely would benefit from additional emphasis on data science methods. Some thoughts.

(1) Are you able to identify some machine learning or other data science methods that IF used in the context of your research and location would present promise towards better understanding your problem? Perhaps a random forest approach from world pop could be used to spatially locate households and persons? Perhaps a hierarchical bayesian model (also from world pop) could be used to describe demographic and economic attributes of persons and households? What about an agent-based model? How would such "what if" applications add value to your investigation?

You might find a few here.

https://scholar.google.com/citations?user=3qWNbcEAAAAJ&hl=en

Even if the study is not exactly Rio, you can still think about how to apply a method in order to surmise how it would advance understanding.

(2) Are you able to identify a gap in the literature? Are you able to formulate a research question that aligns with that gap? Have you classified that research question as a type of "puzzle" that you are attempting to solve?

Just keep going!

good

good

Great stuff. Some thoughts.

(1) You could probably go two different directions with this investigation. I think a lot of the current research regarding malaria and vector borne diseases revolve around using CDR data with gravity models, radiation models and impedance models. You may also find an agent-based model of transportation systems and human movement that incorporates predictions of malaria prevalence. It could also be possible to focus on the natural environmental raster data (precipitation, water, climate factors) as predictors in a machine learning method. While spread of the disease can be predicted by modeling transport behavior, at its origin, most countries that have eradicated malaria and similar types of vector borne diseases have done so through public works efforts that seek out to spray insecticides that kill mosquito larvae in standing water as breeding areas. If it is possible to more accurately (spatially) and then effectively identify these source locations, potentially there could be great progress towards eradicating the disease.

(2) I have also seen a few studies that seek to model the mosquito population itself. I think the limitation of that approach has been largely computational due to the sheer numbers, the scale of the insect and the environmental and natural factors associated. Maybe that has recently changed.

(3) I still wonder if climate change is increasing malaria disease prevalence. Are the mosquitoes and parasites manifesting in locations across Africa where they previously hadn't been observed? Are they more prevalent in China? In other locations throughout the globe?

Just keep going!

good

good

good

good

very good

good, the "night-biting" aspect is often lost on investigators -- has that been definitively proven?

Excellent work. Some thoughts.

(1) You have identified a wide array of potentially viable and effective paths forward. Some of the methods are based more in traditional statistics (poisson, SIR, multinomial logistic regression) but still present promise, especially if they are leveraged with a learning algorithm (MNLR) or possibly a process model (poisson). I'm also wondering about SIR models for simulating an infectious outbreak (using ODEs or an agent-based model).

(2) You might also want to narrow your focus. Taking into consideration where most research is currently "placing its bets" is not a bad way to asses how to best invest your time moving forward. I suggest continuing to focus on CDR data, gravity models (and their variants), radiation models, impedance models, use of voronoi polygons and possibly measures such as radius of gyration (although this may be more effective in the context of a response to a natural disaster). What is state of the art for using CDR data to describe, analyze, model and then predict human movement during an infectious disease outbreak.

(3) I think you could continue to focus on describing the process itself. Why is it important to model human behavior using cell phone data? How does understanding human movement help us to better address an outbreak? How could this information be used in a time of crisis to expedite solving the problem?

Just keep going!

ok, what kind of assessment do you give this source?

good

excellent, I wonder why?

excellent

good

good

good

good

good

good

interesting

good, describes process

good

very good -- perhaps also how it is spreading?

excellent

Good work. Some thoughts:

(1) Further define your focus in terms of the problem you are seeking to address. Will you investigate water use / scarcity in terms of residential demand? agricultural demand? Are you also thinking about water degradation? Are you mostly interested in potable water? Urban water use? Investing a bit more time with regard to your problem statement will help you to narrow your focus and further articulate your research question.

(2) Are you able to identify some machine learning or other data science methods that IF used in the context of your research and location would present promise towards better understanding your problem? Perhaps a random forest approach from world pop could be used to spatially locate households and persons? Perhaps a hierarchical bayesian model (also from world pop) could be used to describe demographic and economic attributes of persons and households?

(3) Are you able to identify a gap in the literature? Are you able to formulate a research question that aligns with that gap? Have you classified that research question as a type of "puzzle" that you are attempting to solve?

Just keep going!

interesting, I'm wondering if there are other methods that may present more promise in describing, analyzing and predicting water use / water demand?

good, beginning to sound like a general research question

interesting

good for thematic and conceptual purposes

Good work. Some thoughts.

(1) Are you able to identify some machine learning or other data science methods that IF used in the context of your research and location would present promise towards better understanding your problem? Perhaps a random forest approach from world pop could be used to spatially locate households and persons? Perhaps a spatial probability density function could be used to identify boundaries of de facto urban areas? Perhaps a hierarchical bayesian model (also from world pop) could be used to describe demographic and economic attributes of persons and households? What about an agent-based model? How would such "what if" applications add value to your investigation? The last two sources seem to introduce these types of methods. Perhaps you could further articulate details from each of these sources?

(2) Are you able to identify a gap in the literature? Are you able to formulate a research question that aligns with that gap? Have you classified that research question as a type of "puzzle" that you are attempting to solve?

Just keep going!

very good

excellent, could you further describe the model and methodology?

good source

did they present a model?

good

good data source, VIIRS is also used

good data source

ok good

good

OK good. Excellent thematic focus and problem statement. Some thoughts.

(1) Are you able to identify some machine learning or other data science methods that IF used in the context of your research and location would present promise towards better understanding your problem? Perhaps a random forest approach from world pop could be used to spatially locate households and persons? Perhaps a spatial probability density function could be used to identify boundaries of de facto urban areas? Perhaps a hierarchical bayesian model (also from world pop) could be used to describe demographic and economic attributes of persons and households? What about an agent-based model? How would such "what if" applications add value to your investigation?

(2) Are you able to identify a gap in the literature? Are you able to formulate a research question that aligns with that gap? Have you classified that research question as a type of "puzzle" that you are attempting to solve?

Just keep going!

ok good

how did they calculate and identify these hotspots?

my old boss!

my colleague!

could you further elaborate on the method?

Excellent - is this a spatial probability density function (sometimes aka kernel density estimation)?

good - targeting public expenditures

good

ok good

good

very good problem statement -- could you further quantify the significance of these harms within the context of Ethiopia?

OK good. Some thoughts.

(1) Are you able to further identify the data and the variables used with some of these investigations related to nature disaster response in the Caribbean? Are you able to identify names of methods used? names of models used? Some of the methods and models that will likely surface are as follows.

• Voronoi polygons - what are these? why are they relevant to CDR data?
• radius of gyration - what is this? what is it measuring? gravity models and other variants? what does a gravity model describe and predict? How is CDR data used to estimate the origins and destinations of all persons (OD matrix) and how is this OD dataset input into a gravity model to describe movement? Is it also useful to estimate future decisions and movements across an area?

(2) Further specify your topic. Is there a particular type of natural disaster / event that you are investigating? Hurricane? Earthquake? Disease?

(3) Are you able to identify a gap in the literature? Are you able to formulate a research question that aligns with that gap? Have you classified that research question as a type of "puzzle" that you are attempting to solve?

Just keep going!

good - but I think assuming the user opted into sharing their location data through twitter?

good

good. Did they use a model in their research?

good source

edit

OK good. Some thoughts.

(1) Are you able to identify some machine learning or other data science methods that IF used in the context of your research and location would present promise towards better understanding your problem? Perhaps a random forest approach from world pop could be used to spatially locate households and persons? Perhaps a spatial probability density function could be used to identify boundaries of de facto urban areas? Perhaps a hierarchical bayesian model (also from world pop) could be used to describe demographic and economic attributes of persons and households? How would such "what if" applications add value to your investigation?

(2) Can you further articulate your problem statement? Which aspect of air pollution will you focus upon with your continue investigation? Have you identified a gap in the literature? How about a tentative research question?

Just keep going!

ok

does indeed sound interesting

good

good, what is ger density?

good

also contributes to outdoor pollution, but would be good to distinguish indoor pollution as a specific topic

wow! a statement such as this would benefit from a citation

It's not clear to me where you problem statement resides. Please integrate your identified harms, their significance and the inherent nature of your selected topic into the introduction / development topic sections of your literature review.

Excellent annotations. You've covered a lot of ground in content here. Some thoughts.

1. Explicitly identify all of the different model involved with these sources. I expect several different ones were used. Can you rank them in terms of which one or two exhibits the most promise in your estimation?

2. State your problem. Identify the most relevant themes and topics. Describe this as a process in terms of China. Try to identify some areas in the literature that exhibit research gaps, and then begin to articulate a research question. Attenuate that question such that it aligns with your identified gap. Classify the type of "puzzle" or research question you have formulated.

Keep focusing on the data and methods, and also just keep going! Good work!

interesting link to your previous source

great

excellent, sounds similar to Zipf's law

power law / Zipf's law ?

excellent, urban sprawl is a very important theme

good, important development themes

was a model identified and used as part of this study?

what is LBS?

good

interesting, good

another good source

excellent

good, but are you able to make this link active?

good

Good source. For annotation style, Gaughan et al. is typically used when the number of authors is more than 3.

Very clear research focus. Some thoughts.

(1) There might be an argument for visiting some of the traditionally considered foundations of development economics. The first one that comes to mind is Hans Singer and his Post-war Price Relations between Under-developed and Industrialized Countries (1950) and how it contributed to the Singer-Prebisch thesis. Central to this thesis is the idea that due to the terms of trade, developed countries have a moral imperative to return investment (perhaps repair would be a better choice of words) under developed countries from which they have benefitted. You could also look at Sir Arthur Lewis' Nobel Prize winning essay from 1954, Economic Development with Unlimited Supplies of Labor. Both of these sources were instrumental to the foundations of development economics and find corollaries with the current state of under developed regions of that were sources of the transatlantic slave trade. Amartya Sen also provides some insight to this historical perspective on development analysis.

(2) I also can't help but think about the American influence with regard to your topic. While these countries have lost human capital as a result of the transatlantic slave trade, which countries have in fact gained? Can you evaluate this gain somehow and is there in fact an argument in favor of repairing source contributing regions? It's a worthy justification for your investigation and goes straight to the inherent and complex nature of the problem.

(3) How can you take some of the surveys on education and apply a data science / machine learning methodology in order to produce a high resolution, spatial description of education, level of service across West Africa. I suspect many of the big steps forward with progress are going to be at the level of individual schools, throughout Ghana, throughout Nigeria, throughout Liberia and elsewhere in the ECOWAS community. What if you applied a hierarchical bayesian model to some of the survey data you found. Would it be possible to use such a machine learning approach to produce a high resolution description of educational level of service of primary and secondary schools throughout West Africa? Would such a high resolution description of all persons living throughout West Africa as well as the level of educational service they are receiving be useful to inform implementation of policy at the national, district and local level?

All around great stuff. Just keep going.

significant

interesting

How has Liberia fared since their civil war? Have things begun to improve?

good

I'm wondering if he is in favor of advancing indigenous institutions as a means for promoting development. Do you think there is evidence that favors advancing indigenous institutions when compared to modern ones, especially when considered in the context of West Africa (Ghana, Nigeria etc...)

good

ok

good

Super. Clearly defined research focus, although a massive topic in terms of scope and scale. Some thoughts.

(1) Might not be a bad idea to revisit some of the traditional development economic works on rural to urban migration, although the push / pull factors you are identifying are much more regional or global in nature. Two good ones to look at are Todaro:

https://www-jstor-org.proxy.wm.edu/stable/1807860?seq=1#metadata_info_tab_contents

and Sir Arthur Lewis' Economic Development with Unlimited Supplies of Labour. Please let me know if you can't find a copy of his 1954 essay that earned him the Nobel Prize.

(2) I'm wondering a bit about which data science methodology is best to investigate further with regard to your research topic. Argument could be made for thinking about the application of a hierarchical bayesian model with the survey data you have identified. There are also arguments in favor or gravity type models with the CDR data you are considering.

(3) Migration patterns are such a big topic that it can be difficult to know where to start in terms of geospatial data, data science / machine learning methods / models etc…A common starting point for thinking about migration is in terms of push factors and/or pull factors, more traditionally within the context of rural to urban migration. The scope and scale of what you are investigating is of a much greater magnitude and complexity than these traditionally considered foundations from development economics. Additionally, thinking of migration of this magnitude in terms of only human movement and behavior, likely over simplifies the problem. For example, we could use CDR data to infer the origins and destinations of where people are coming from and where they wish to go as part of a transportation system model (even if it is by foot or mixed mode). I think you need something that seeks to accomplish far more than simply transport behavior and movement, because you also wish to understand the motivations behind why people choose to leave (push factors) as well as the reasons they are attracted to some location (pull factors). I think looking into agent based models as a data science method for describing, analyzing could also be a promising path forward. A simulation system, such as an ABM, may incorporate different model that address factors at the origin as well as at the destination.

Will be very interested to find out more about the direction you have selected moving forwards! Just keep going!

excellent

good

PUMS is a good data source

How did they go about doing this?

good

very good

interesting, was a model used that incorporated both the household surveys as well as the geospatial climate and topology data?

good, DHS? living standard surveys?

Very good, particularly your classification in terms of voluntary migration. Have you had a chance to consider push / pull factors?

Nice link!

Very good work. Some thoughts.

(1) I'm wondering a bit about which dimension of health care / disease you may be focusing upon in relation to Tajikistan. Your problem statement indicates you are interested to investigate respiratory health but your annotations do not seem to provide that much information in support of this direction. Perhaps further clarify which dimension of health care you are interested to investigate as part of your literature review. A common approach is start with either infectious diseases, vector borne diseases, non-communicable diseases or psychological disorders. I think you are interested in non-communicable diseases (respiratory / pulmonary, circulatory / heart) and the factors contributing to their prevalence in Tajikistan.

(2) As you have noted, it's likely going to be difficult to find a perfectly matched source to the problem you have presented. Still, you can consider what if scenarios, in conjunction with geospatial data that is available in other locations and contexts as well as the applicable method. For example, what if the authors used a hierarchical bayesian model to spatially describe, analyze and predict outcomes in the context of a particularly selected variable (as well as the predictors). It's one option forward, I'm sure there are many others as well.

Otherwise, just keep going!

excellent

good

was location of health care facility included in the analysis?

good

Is there a spatial dimension to this analysis?

what are these distances measuring?

ok interesting

is this DHS data?

interesting, I didn't know this at all

good

Great stuff. Research focus is clearly defined. Some thoughts.

(1) You have many economic and demographic elements to consider through your investigation. I think you likely want to start with the survey data and the hierarchical bayesian model. This will probably move you in the right direction of better understanding your problem in terms of scope and scale. Your issue as you progress is identifying how your topic intersects with the larger development theme you are considering (wasn't there another dimension you were investigating as well?)...Perhaps start with the survey, machine learning / hierarchical bayesian model approach and consider what data is out there that could be used in the same manner. Even if you don't find a study that is specific to your location, you can still hypothesize how data or a model from a particular study could be improved upon with an improved methodology.

(2) I also want to encourage you to explore what else is out there, beyond hierarchical bayesian models / machine learning methods. Are there any other machine learning approaches that could be useful towards spatial description, analysis and prediction of your development problem. What about deep learning approaches? Is there anything new on the horizon that presents great hope and promise towards a fuller understanding of a large, complex problem from a national to regional scale?

Other than that, just keep going!