10/15: Andrew Stier and Marc Berman

[shevajia]

Comment below with questions or thoughts about the reading for this week's workshop.

Please make your comments by Wednesday 11:59 PM, and upvote at least five of your peers' comments on Thursday prior to the workshop. You need to use 'thumbs-up' for your reactions to count towards 'top comments,' but you can use other emojis on top of the thumbs up.

[bakerwho]

Thanks for sharing this research, and big shoutout to our hero Nak Won Rim!

I was very intrigued by your counter-intuitive result and impressed by the multiplicity of your skepticism for your own claim. Particularly with the Tweet data, the use of the logistic regression to determine if depressed users had different Tweeting habits was commendable.

The physicist in me has taken the bait, and I can't help but wonder what dynamics you might have observed in the BRFSS data (considering it has longitudinal information). I'm curious if we can expect a 'middling out' effect here - where the largest and smallest cities might be better equipped for growth and expansion than the ones in the middle. I'd guess this might be the case if larger cities are better at economies of scale, while smaller ones might have more space/flexibility for sustainable expansion strategies.

Consequently, the 'choices along varied dimensions of life' (such as connectivity, etc.) might be better at keeping pace with the growing population in some cities than in others. And depression rates might show periods of rapid growth in the latter case. Did you observe any such (or other) dynamics?

[william-wei-zhu]

Thank you very much for the interesting paper. The paper suggests that an increase in a person's number of connections causes the likelihood of depression to fall. I wonder if it's possible for the relationship between the number of social connections and depression to be spurious (correlation rather than causation). For example, an alternative explanation can be that big cities have a greater number of interesting events going on than smaller cities. A greater number of interesting events causes both the number of connections to grow and depression to fall. How does your research reject this alternative explanation?

[wanitchayap]

To what extent do you think we can generalize this result to other geolocations beyond the US? The paper briefly mentioned that evidence for urbanization and depression are often inconsistent and mixed partly because of differences in definitions of what constitutes urban. From my own experiences in Thailand vs in the US, the contrast of urban vs rural (or even smaller urban vs bigger urban) is quite different between the two countries, and I think this is probably true for many other countries too, especially if we pair developed vs developing countries. And kind of a tangent to @bakerwho 's note on city growth, I think that most of the US metropolitans are already urban and more established. What about cities in other countries (probably more on developing ones) that are actually in the step of forming (and then there would be differences in the pace of urbanization too)? How could we modify this framework to use with such cases? What do you think are the needed measurements of urban beyond city size so that we can address such cases?

Ps. 👏 👏 Nak 👍 👍 Won 💯 💯 Rim 🎉 🎉 So proud of you!

[nwrim]

Hi Andrew and Marc,

When I was working with the twitter dataset and sought help from a friend on an issue regarding python2, he did not really understand how we can "detect" depression from texts, saying that "Isn't depression in a continuous spectrum? Where do you draw the line between depressed and not depressed". Indeed, our model treats depression as binary outcomes following a binomial distribution, and I did not raise this point when we were discussing the paper because I think most psychologists would not really oppose to this assumption considering our long tradition of diagnosis (with being detrimental to everyday life as a major criterion). But do you think we could adjust the model to treat depression as a continuous state, rather than binary? I know that our data does not really fit when we go down this route but I am curious about your opinions.

Also, I read Omid's 2015 paper on the effect of neighborhood greenspace on health perception/cardio-metabolic conditions (but insignificant effect on mental disorders) a few weeks ago. Looking at this paper after reading this, I wonder if we can do a similar thing with greenspace data in the cities with the datasets we used for this research.

[Raychanan]

In this paper, which challenges conventional wisdom, your results show that depression is less prevalent in large cities and you validate this result with three different datasets.

I'm curious how you see the impact of the COVID-19 on the relationship between big cities and mental health? Do you think the city still plays the role of a "buffer zone" against the epidemic? Or is it that cities are bad for mental health, as the conventional wisdom would have you believe.

For what I have seen in this epidemic is that urban populations have better and more timely access to information than those in rural areas, but the result is deeper emotional damage and greater suffering.

Thanks!

[JadeBenson]

Thank you for sharing! This is an interesting and very relevant paper.

I'm wondering - how do you separate meaningful depression from humor on Twitter?

Online communication is so ironic, uses nihilistic humor, and jokes about depression - how do you distinguish between this and depression that impacts people's quality of life? Do you consider this humor to be a symptom of depression or something separate? Depending on your theory of depression and humor, how would you update your methodology to reflect that?

[TwoCentimetre]

I am confused why you set all the modules in this paper in a nonlinear way. And why demographic or cultural differences are not taken into consideration. To be honest, I do not fully understand this paper. And I am looking forward to your sharing in the workshop.

[yierrr]

Thanks for such an intriguing paper. I am wondering about the representativeness of the data—despite that there are three data sets combined together, but data from all three sources seems to be self-reported (for the Twitter case,even though they are passive observations, Twitter users still have to volunteer to post something in the beginning). Meanwhile, I think there are statments on how depression would bring patients loss of interests as well as motivation to do anything, and in that sense actual depressed patients might be less inclined to report any data in any way, including posting on Twitter. Therefore, would it be possible that the study is actually investigating people who are sad instead of being actually depressed? Thank you!

[bowen-w-zheng]

Thanks for sharing this research. Very interesting and surprising results. For the Twitter dataset, I am really surprised at how consistent the estimation is with results from the other two datasets. If I understand correctly, the dependent variable is calculated with the prediction of people's mental state, which is effectively equivalent to observing the dependent variables with substantial measurement errors. This seems to be pretty common in computational social science.

I am curious about how to interpret estimations obtained from data whose dependent variables have non-trivial measurement errors (e.g. outcomes that are predicted with moderate accuracy).
Does it converge to the true estimation? What are the caveats of doing inference or estimation with such data?

[ginxzheng]

Thank you for working on depression issues and I find it very practical and insightful!

One question is, would it be possible to share with us the examples of depression-related tweets texts? Personally I really love the topic models approaches, which can reveal many angles of discourse. I was wondering the reason you chose to emphasize on tweets count, and will you have further investigation in the 9 different topics specifically?

On the other hand, in terms of psychological symptoms, especially depression, I know sometimes when it's getting severe, depression will deprive you of abilities to talk and even express yourself. With regards to its correlation with city size, would it be possible to address more on those underrepresented groups of people who are less able to let people discover their symptoms?

Another thing I was unfamiliar with is the useful indicators in urban related research. I am wondering is there any better parameters to illustrate the urban dynamics? For example, density, or social distance between people? Do you think, and how do the survey data and geo-twitter data can be transformed to calculate more precise measures?

Thank you!

[ydeng117]

Thanks for sharing your research. I am curious about how you consider the geographic setting for cities. In America, people tend to live in suburban or in rural places, while in many European and Asian cities, people often living in urban areas. I wonder how would this difference in residential preferences affect your results. Also, I am wondering about the differences between the major metropolis and small cities. Thank you!

[heathercchen]

Thank you so much for sharing this wonderful paper! The topic is quite interesting, especially for the results, you get one that is against our conventional thinking. My question is about how to manage the socio-economic controls. It seems to me that it requires a series of socio-economic control variables to get the casual relationship between city size and depression. Could you please explain the criterion that you choose these socio-economic control variables and how to make them consistent across different datasets? Thanks!

[wanxii]

I, myself, am not quite a 'fan of city'. So, I might be 'critical' of your result.

1. I think building a model should be a far more rigorous task, compared to empirical studies. Personally speaking, I think the model in the paper might be a bit simple and blunt, so it might only explain the correlationship (as @William has mentioned above). However, it also means that there would be lots of interesting follow-up studies (i.e. building a more complicated model to explain the causality). Therefore, look forward to your further researches! But before that, I think there's still one important thing requiring concerns: why the model fits so well in big databases despite its simplicity? Or maybe the databases are biased themselves, and hence their properties accidentally suit the model.

I haven't had a clear picture in my mind yet, but I'm just thinking from whom we collected the data. Whether it's NSDVH or BRFSS or Tweets, we're collecting our data from people who still have connections to the world and their behavior paterns might fit the model. But what about people we can't 'see' in the datasets? What if people with MDD (who neither speak up online or offline) mainly reside in larger cities? What if those 'cold', crowded big cities breed silent sad people and its convenience makes it easier for them to hide in their own isolated tiny space? Therefore, the percentage of people in depression in large cities might be underestimated. I wonder what's your opinion towards it. (I saw @nwrim's note below and I totally agree with him, so I've elaborated my concern.)

2. Just a trival advice: I think it might be better if starting with descriptive data and simple regression analysis after making the assumption, rather than going straight to the model. If the results derived from the raw data align with the assumption, then we can smoothly introduce our model. If not, we could propose it's because of the mediation effects of social networks and go on examining the model.

[linghui-wu]

Very interesting paper, though I am somewhat suspicious of the model setting and data collecting. I agree with @heathercchen that socioeconomic determinants such as the advanced medical infrastructures in the metropolis may play an indispensable role in alleviating depression rates. On the other hand, individuals diagnosed with depression tend to isolate themselves from the outside world, as mentioned by @wanxii, which probably renders selection bias in twitter data.

[YuxinNg]

Thanks for the interesting paper. I am curious about if the result may change given another area (let's say another country) and another period of time (e.g. the Great Depression or during the Financial Crisis). It may be interesting to pursue. Thanks.

[mikepackard415]

Thank you for sharing this exciting and important research. As Geoffrey West describes in his book Scale, which I expect you're familiar with, cities display superlinear scaling relationships with all sorts of measures related to socioeconomic connectivity, and given the theory that inversely relates connectivity to depression, it makes sense that depression would scale sublinearly with city size.

The last chapter of West's book discusses the long-term sustainability of super-linear growth in cities, and how it requires ever-accelerating innovations and paradigm shifts in order to avoid stagnation and collapse. My question is this: In your view, do these results about how depression scales with city size change the prospects for whether sustainable cities are possible? Maybe happier people innovate more? Thanks again.

[PAHADRIANUS]

I think the paper offers a solid and convincing proof of the connection between cities' sizes and the mental health of their residents. Still subtle distinctions should be made between city dwelling and urbanization. As put forward by the paper's introduction, differences do exist between rural and urban regions and urbanization is generally positively correlated with depression. Thus when looking into a specific region, it is tricky to identify where the boundary of the city lies. The surveys that supply data are made with random correspondents based on administrative divisions and thus do not detect their exact residential environment. A respondent identified as a resident of a large city may live in a neighborhood that is not sufficiently urbanized as metros such as NY or Chicago do contain rural areas. I guess an improvement could be instead of using cities, which vary greatly in many aspects and therefore are hard to standardize, to compose main explanatory variable, using randomly sampled regions of a certain size supported by their urbanization characteristics such as population density, transport and public good infrastructure, or migration dynamics.

[tianyueniu]

Hi Andrew and Marc,

Thank you for sharing this research at our work shop! This is a really novel way of looking at urbanization and depression. I personally find the results intuitively convincing as I do think more social stimulation/distraction/support would make people less prone to depression. I wonder if further stratification by 'years lived in a city' would bring more insights into the analysis? New-comers might have less exposure to the variety of social stimulations available in a large city. I think this information might be readily extract-able from tweets with geo-location information.

Another detail is that, in the paper it was mentioned that a previously defined lexicon of seed is used to organize tweets to correspond to topics in PHQ-9 to guide LDA, I would be very interested in knowing which lexicon is this. Did you examine the precision of this model on your own? Do you think it would be worth it to build a new lexicon to achieve higher accuracy in identifying depression?

[harryx113]

Thank you so much for the presentation! Following @ydeng117 ’s question, the connotation of living in a city vs living in suburb can be very different in different countries. For example in more populated countries, living in suburb can be a luxury which means fresher air, cleaner water, and greener landscape. It would be interesting to see comparisons between cities to further understand the relationship between urbanization and depression.

[lihanhuisherry]

Thanks so much for sharing the paper with us, Andrew and Marc! I think it really interesting to see such an innovative view on the relationship between urbanization and depression.

I am very interested in the content analysis component of this research and I would definitely love to hear about the quantitative methods that you used to assess the colors/emotions of the tweets during the workshop. Thanks!

[wu-yt]

Thanks for the interesting paper. I find this result very counterintuitive because I believe that cities are detrimental to mental health. How would COVID-19 crisis influence your research?

[a-bosko]

Thank you very much for your research! I find this area of study incredibly fascinating and I am happy to see that more research on mental health is being conducted.

The article states that there is a difference in depression risk based on the density of the population. The more dense a city is, the lower the risk for depression for the average person. Is it possible that since there are more opportunities to create social connections in more dense cities, people are less anxious about being kicked out of a social group or social pairing? For example, if an individual has 20 connections, they are less likely to care if one of their social connections breaks off. On the other hand, if an individual only has 3 connections, they may feel more anxious about being alone.

Because of this thought process, is it possible that loneliness can increase or decrease the risk of depression as well? Since humans are social creatures and depend on other humans to survive, could there be an underlying mechanism of survivability?

Thank you, Angie

[Dxu1]

Thank you for sharing the paper! I have two quick thoughts/questions:

1. In the paper you discussed the choice of data (survey vs medical record). How would you evaluate the reliability of survey data vs medical record, considering in survey data respondents may have a more "inaccurate" assessment of their mental state than a doctor? How do you think using medical record data (if possible) would change your results?
2. Out of the three datasets used, it seems only BRFSS reports gender and parameters related to social-economic status. I am curious if you have find the relationship between mental health and city size differ along the dimensions of gender/race/social-economic status (if sufficient data was available.)

Thank you very much!

[SoyBison]

It's clear that Nak Won Rim is dominating our workshop this quarter and I for one embrace our Nak Won Overlord.

It seems like a lot of people really don't want to accept your conclusion that the popular intuition about happiness cities and rural areas is wrong. Personally, I've lived in a lot of different places, both rural and urban (all in the US) and I can say for certain that it is not as simple as that. I think papers like this that challenge our intuitions are the backbone of science. There are few philosophers of science that would agree with the statement "Science exists to prove our intuitions right."

All that being said, in doing this research, did you run into any such challenges? Overcoming internal scientific bias is, in my opinion, the most important thing that a scientist can do. Over the year that I've been in this program I'd say that the thing which has disappointed me the most about the social science community is the willingness for social scientists to justify their agenda as moral imperative, and allowing those intuitions seep into their work.

To reiterate my question, what do you think about all this? Did you run into any such challenges during this research? I've heard through the proverbial grape-vine that some of the co-authors may have had to confront this during this research.

[kthomas14]

Thank you for sharing your research! I see that the model simulation was generated using the socioeconomic networks of individuals and this was because people who live in urban areas are more likely to have exposure to all different levels of socioeconomic status. My question is, were there any other indicators of large social networks that you were considering when building the model? If so, why did you decide on socioeconomic status?

[qishenfu1]

Thank you for sharing your interesting work! First, I want to say that I strongly agree with your research result because I have a similar personal experience. I did my undergraduate in a large financial center in the world. During my junior year, I went to a town in California for exchange studies. I found that my mental status is better during normal undergraduate studies than during exchange. Because in a large city, the transportation is very convenient, and I have more channels for entertainment. In small towns, it is even difficult for me to get from one place to another because there is almost no public transportation. Second, your research design and simulation is rigorous and coherent. I really enjoy reading your paper.

[jkatz913]

It seems like, to date, there is mixed evidence in terms of the relationship between city size and mental health. Is that a potential indication that the nature of this relationship depends on the city, on the population being studied, and the variables or survey used to operationalize mental health? My question speaks more broadly to the pursuit of universal theories in the social sciences. I would be curious to hear your thoughts on this tendency in the social sciences of establishing universalist theories.

[siruizhou]

Thanks for sharing this work! Unlike @wanxii, I am a big fan of city so I'm also gonna be critical about your work.

I don't think this counter-intuitive result is solid and meaningful enough.

1. I think urban density is most closely related to economic activities. More than 80% of global GDP is generated in cities and roughly one-third of global GDP is generated within the 123 largest metropolitan areas. I think it is necessary to consider the economic status and income inequality when examining mental health issues.

2. As @william-zhu-3541 mentioned, cities have better entertainment and cultural life. I could come up with more such hypotheses, such as better services and amenities, which could be the buffer to depression and also are highly correlated population. Big cities are indeed attractive, especially to high-skilled workers. Maybe it is more about who you interact with than how many people you are interacting with.

3. Your results suggest "larger city environments and urbanization may, on average, actually provide greater social stimulation and connections that may buffer against depression" Could it be the case that larger cities are less tolerant to depression and people actually have to relocate? Is there a mental health sorting effect?

4. I wonder if it is meaningful to analyze depression at an aggregate level. Empirically, “happy families are all alike; every unhappy family is unhappy in its own way” said Leo Tolstoy. I bet it's quite hard to get any conclusive result at the urban level. So could you explain why you think it's also important to investigate the urban setting as a whole, compared with on contextual and environmental factors?

Just as you mentioned in your work, (and @wanxii mentioned earlier), there are many follow up researches and I'm very excited to learn more!

[NikkiTing]

Thanks for sharing your work! The results are very interesting. Although, as mentioned by many others, there are still a number of important factors that seem to have been left out of the study. I think it will be really worthwhile to revisit these.

While the paper focuses on the difference between city sizes and the volume of social interactions based on these, I was wondering if structural factors (i.e., infrastructure) that differ between cities can also be considered in terms of doing research on mental health and urbanization.

Also, I am interested to know your thoughts on what you may think your results would mean for cities in developing economies. I personally would expect the opposite result.

[rkcatipon]

Thank you, Dr. Stier and Dr. Berman for sharing your work, I look forward to your presentation! I appreciated your paper's willingness to engage with the limitations of Twitter as a representative data source. This is less a question but more a musing on the modalities of Twitter as a space for identity and belonging, much like cities are shared spaces: I wonder if our online digital communities in any way supersede our physical locations? For example, a popular community like Black Twitter might be considered an ethnic enclave of the internet, much like Chinatown in SF was once an ethnic enclave for Asian immigrants. Therefore, could depression also be examined as a function of more ephemeral digital locations?

[Qlei23]

Thank you for sharing your work with us. I have the following concerns:

1. I think the model construction is a bit arbitrary. Is there any empirical basis for the nonlinear specification & probability function/distribution?
2. The model is likely to capture the correlation, instead of causal relationship between depression rate and population size.
3. I'm also curious about the text analysis and why you choose Twitter data in 2010 (a year when people have not started overusing social media).

[Bin-ary-Li]

Thank you for the stellar paper. I really appreciate the cross-disciplinary approach in used here and I think it is brilliant to combine methodologies used in different fields from machine learning to social network analysis. My questions are:

1. I have qualm about the validity of using Latent Diritchlet Allocation on the twitter messages to tag "depressive" individual. How accurate/effective is that method? Is there previous work on that?
2. I did not to find the connection being made between the analyses of the three dataset and the depression-social network model Y(Ni, t). I am sure it is somewhere in the text but maybe it can be emphasized a little bit more?

-BIN LI

[Yilun0221]

This is really an inspiring paper! Actually it is the first time that I read papers that link urban science to psychology. It is really important to measure how a city is comfortable for people to live. However， I wonder whether it will cause omitted variable bias if we only consider the size of a city?

[MengChenC]

Thank you for the amazing work. Both the research design, from data collection to cross-examination, and the outcome are appealing. I am interested in the part you adopt logistic regression to distinguish individuals who have depressive symptoms from their tweet frequency. Would you be able to talk more regarding the process? (e.g. How to decide the specified number of tweets for cutoffs from 0 tweets to 110 tweets, etc.) Thank you.

[MkramerPsych]

Professors,

Thank you for sharing your work with us! This is incredibly rigorous and applicable research. I wanted to ask:

1) Regarding the operationalization of depression, I noticed that the metrics included the use of the DSM definition of depression as an indicator. In addition, the model only predicts depression through the use of a binary variable representing whether or not depression manifested in the person based on the previous metrics. Is there a way your model could examine depression operationalized by its symptoms directly? I am curious if varying levels of depression can be tied in any substantive way to geographic location/urban vs rural environment or to the relative size of the individual's socioeconomic network.

2) In terms of the use of Machine Learning to extract depression symptoms from twitter data, what procedures did you/can you employ to make sure that sentiment is not confounding your results? Isn't it entirely possible that people are putting up a facade on social media rather than being entirely honest? Also, is there an effect size issue in the twitter data? I have a feeling that many more twitter users are located in urban areas than in rural areas.

[chiayunc]

Thank you for sharing your wonderful work. In the early movement of urbanization, many argued that urbanization cutoff social connection. I think Jane Jacob would be a great example. If you were to get a hold of datasets that would cover the shift from rural-based lifestyle to urbanization, do you expect to see similar results? Do you think that adapting to existing cities and changing the living environment at the same location from rural to urban might be fundamentally different?

[alevi98]

Thanks so much for a thought-provoking paper! I had a few follow up questions/points. I understand the intuition for running an LDA analysis on large-scale Twitter data to detect "depressed" language. However, I'm wondering if there's any further justification for generalizing this finding to the population at large? My understanding is that there's a correlation between social media usage and feelings of depression. If your data was all from social media, were you worried about multicollinearity?

Moreover, did you think to account for disparities in intergenerational usage of social media? For one thing, people on social media are disproportionately younger, but also the conventional wisdom is that as we get older, we like to move away from the city & into the suburbs to raise a family or the deep wilderness to fully detach from our material possessions. Do you think your result is generalizable to all age groups?

Another more philosophical point is about treating "depression" as a binary variable based on the DSM manual. There is an emerging body of literature in psychology about whether or not we should even have a DSM because it is too prescriptive and does not account for how definitions of mental health vary by culture. How does this idea impact the scope of your findings?

I'm excited to learn more. Thanks again!

[hihowme]

Thanks a lot for your paper! It's really interesting. There are also lots of interactions between psychology and economics, especially in the field of behavioral economics and consumer behavior. I am wondering is there any implications of your study that the economics or business study could borrow from? Thanks a lot.

[Yutong0828]

Thanks for this interesting paper! I found the results quite amazing and even a bit counter-intuitive, as it's natural for us to think that city life may cause more pressure and the relationships among people are less close than that in the countryside.

In your study, it seems that you compared the statistical data of different cities, and proved that larger city will actually have lower depression rate, but I was wondering if we could further infer that people in large cities are less likely to be depressed than people in rural areas or countryside? Because according to your work, larger city is more beneficial because there are denser social networks, and if this is a decisive factor, we should get similar results when comparing between cities and countrysides.

However, as there are also risk factors which could overshadow the benefits brought by social networks in the cities, it's still hard to reach a conclusion, and I have noticed that for smaller cities you didn't get significant results. So, can we say that smaller cities have higher depression rates than both of the larger cities and rural areas?

Thanks a lot!

[AlexPrizzy]

I find this approach to studying health impacts of urban living through analysis of social space and health rates to be very interesting. Though I can imagine this research being further expanded to look at more specific subgroups of people based on individual differences. The correlation of depression and social interactions can drastically vary between individuals, ex. an introverted person may suffer from large degrees of social connections while an extroverted person may benefit from social interactions. Suggesting "stimulus overload" as a possible negative aspect of urban life may not be true for all either, individuals with hyperactive disorders may flourish in such an environment as research suggests these individuals may be better at navigating complex environments. How could we ethically collect data on such subgroups?

[egemenpamukcu]

Thank you for sharing your work, it was a very thought provoking read. Although at first glance it seems counter-intuitive that urbanization can lead to lower rates of depression, essentially haven’t humans built cities just for the reason they build almost anything, to solve their problems? The biggest impact of cities, labor specialization, allowed unprecedented growth in economic productivity as well as leisure time both of which must have very close ties to depression. So on a second thought I am not very surprised by the findings although I might be a bit biased as a vocal critic of the whole "return to the nature" and "back to our roots" bandwagon.

So, I was wondering, do you see this relationship changing in the future? With improved communication capabilities presented to us through digitization, some seem to think that the end of "the city" is near. They argue that one of the very reasons for the existence of cities, the need for workers' physical proximity for the production of goods and services, is now fading away and this could eventually lead to a more even distribution of population. However, it is unlikely that the number of interaction people have will decrease, those interactions will just move to the digital domain. Do you think this changing trend in the way we interact can tilt this negative relationship between the number of social ties and depression? Or does digitization, by allowing nonurban populations to have more interactions, can help with anxiety in rural areas?

Additionally, compared to the country, cities have much more developed infrastructure and skilled personnel to help people cope with mental health issues such as depression. How big of an impact do you think the improved access to mental health resources have in this relationship? Or is it just that they have more access to just about everything (entertainment, education, etc.)?

[LFShan]

I was surprised by the result. I used to think that people who live in the city should experience more pressure because of their limited living space and highly competitive job markets. Would the rapid development of communication technology change the current situation in the future?

[ghost]

What do you think is the reason for the sub-linear relationship between depression and city size?

[hhx2207061197]

Thanks for sharing such an interesting paper! I'm wondering how sudden global diseases such as the COVID-19 affect your research results?

Thank you!

[chrismaurice0]

Great article! Excited to hear you discuss it tomorrow.

My question is on the decision to use city size. I could not help but think about the differences of living in Chicago compared to Houston, two cities with similar populations, different densities, and different ways of life. For example, traveling 5 miles in Chicago is different than traveling 5 miles in Houston. Cities may be the same size in population, by structured very differently. Could those differences in structure lead to increased/decrease rates of depression? You state k0 is a prefactor independent of city size, and ξ a residual measuring the distance from the population average. What if you added spatial data into your model that included not just in distance but average travel time from the entire city to a population centroid?

[NaiyuJ]

Thanks for sharing your work! The finding that larger cities show substantially lower per-capita rates of depression does not seem to make intuitive sense. People in urban areas are generally under more pressure from work, life and personal finance. The diversity of urban residents in terms of their social status, financial wellbeing and living conditions all contribute to their mental health. Measuring the per-capita rates of depression is unable to reflect the underlying diversity nor can it do justice to pockets of the population with high depression rate.

[yutianlai]

Thanks for sharing your work! I'm wondering how you differentiate the effects of other socio-economic factors?

[jsoll1]

Thanks for sharing this paper with us!

I agree with the people questioning the usefulness of twitter analysis as a way to calculate mental health breakdowns over a region. I have a question along William's train of thought: do you think that maybe an alternate explanation is that therapists are more easily accessible in cities, which could result in lowered depression? Or could people's living situations (e.g. in a city or not in a city) impact their decision of whether or not to spend time tweeting?

[Anqi-Zhou]

Thanks for the sharing! It's really an exciting work. My question is how you can quantify/estimate the emotion in the tweets? We'd like to hear more details about this question. And the answer be really helpful to guide our own research work.

[ddlee19]

If the rate of urbanization is rising on a world-scale, is it possible that more people living in urban areas feel "less depressed" because the large majority of one's social circle also live in urban areas and their relative happiness levels are comparable although their absolute happiness level compared to someone living in rural areas is quite low?