Making Spatial Weights: Part 2

We create many weights matrices and use them to compute spatial correlations, and see if different ones matter. You can save these to use in R, MatLab, Stata, python, ... Geoda has many tools to make, save, and visualize your weights matrices!
Download File: www.google.com...
Spatial Video Playlist: • Spatial Econometrics (...
Spatial Stats Website: spatial.burkey...
Link to video on "Great Circle Distance" Formulas versus Euclidean: • Great Circle Distances...
My Website: www.burkeyacade...
Support me on Patreon! / burkeyacademy
Buy me a cup of coffee on PayPal? paypal.me/Burke...
Talk to me on my SubReddit: / burkeyacademy

[@maxfancourt8347]

Just came across this series of videos, and can I say just how refreshing it is to find clear, well thoughtout and complete explanations of spatial statistics. You are a great teacher, and you have made this topic very accessible!

[@BurkeyAcademy]

Thanks for the feedback! Glad to have helped!

[@amandabotticello2722]

Thank you so much for your videos. Everything is well explained and I appreciate your two cents when giving advice on what approach to use for creating weights.

[@vaibhavpuri9278]

Very Well structured and explained. If you could please add more videos on GeoDa.

[@farahnaz2384]

what is FIPS,Why it is ID variable??i created an ID variable, it says "poly ID", is it good to go?? i need a response on it urgent plz

[@BurkeyAcademy]

@@farahnaz2384 In the US, FIPS is how the government mapping people assign IDs to regions. Yes, you can create your own.

[@faysalkabirshuvo737]

Great series of video. Just wondering what is the conclusion about using a distance weight matrix? which option we should use from queen, rook, distance and KNN?

[@shamuhammethajyyev3336]

Thank you for the video, why do u choose fips all the time and what does it mean, thanks beforehands for response.

[@BurkeyAcademy]

FIPS is the US stands for Federal Information Processing Standard, and each state, county, and census tract has a standard ID number.

[@JacquesHenriqueDias]

Thanks for the video it was very useful again! A question, if I'm dealing with coordinates, latitude and longitude, is it better then to use k-nearest weights matrices or a distance based bandwidth... I am working with adresses in urban areas.

[@zahramovahedinia1896]

For me the values in the .gwt file is not between 0 and 1.

[@ProfBoggs]

1. Virginia is weird. That city independent of a county administrative system is the opposite of a unigov (e.g., Indianapolis in the late 1960/early 1970s expanded its bounds to subsume all of Marian County). 2. Does Geoda allow you to exclude polygons if the polygon has no neighbors?

[@BurkeyAcademy]

Virginia is a real pain, especially since sometimes the Independent Cities are absorbed by the county, making comparisons over time a painful exercise. I don't use GeoDa much anymore, so I'm not certain.

[@ikik5511]

Sir, I want to ask you a question. Sometimes we have "island" problem bcs of a missing data for spesific regions. For example, now I have a data set for two different region and each region consists of at least two cities. However, since the lack of data set of the area btw 2 regions, I have 2 different big island and both island has at least two city. Can I form weight matrix for these dataset or should I find the missing dataset btw these two regions?

[@BurkeyAcademy]

I have not personally dealt with this problem, and am not comfortable working with missing data. However, there are some papers where people have worked on this, for example: www.sciencedirect.com/science/article/pii/S2213398420302232 , www.semanticscholar.org/paper/Missing-Data-In-Spatial-Regression-1-Boehmke/528d33795c0a9fb3d9a7cad97f1ae7e31020b5d0

[@ikik5511]

@@BurkeyAcademy Thank you so much! Thanks to your website and other resources that you shared with us, I improved myself.

[@SuperAdeoluwa]

Hi, I need to import a gwt file into stata, but I get the following error : "spwmatrix_CalcSPweightM(): 3200 conformability error : - function returned error" Do you have any suggestions that can help fix this problem? I've been using "spwmatrix import using "C:\Users\adeol\Desktop\test.gwt", wname(weights) xport(weights,txt)" to import the gwt file into Stata?

[@BurkeyAcademy]

Sorry, I don't use STATA... hopefully if you post in a STATA forum someone can help figure it out. Best of luck!

[@deanolson9600]

Regarding the distance values you are encountering around 26:03 - are these perhaps distance in arc-second format? Given that this data was originally in an unprojected format (WGS 84) it seems likely that this is the internal linear unit. I have no idea, but it's a suggestion. Great courses!

[@BurkeyAcademy]

Thanks for the idea- appreciate the feedback!

[@tawilk]

he used Euclidian distance for the 2 knn matrix. so the distances are being reported in the generic (inch?" distance b/w the centroids on the map.

[@jane7225]

Nice video!!! I want to ask if it's ok to use binominal variable(0,1) with the same method you did when making spatial weight matrix .

[@BurkeyAcademy]

Are you saying that your dependent variable is binary? If so, yes. It is possible to estimate spatial probit models with a binary dependent variable.

[@jane7225]

@@BurkeyAcademy yes. I meant dependent variable. Is the way of making spatial weight matrix and computing moran's i same as you did?

[@naruyah]

good

[@Pepelao1295]

same problem with the weight maanager!

[@BurkeyAcademy]

Thanks for letting me know- I guess restarting is what you have to do...

[@josepmarti2025]

Nice videos about GeoDa and how to use it. I just have one question. I would like to develop a non-spatial weights matrix, for example with gdppc. Is it possible to it with GeoDa? and If yes, could you explain me how to do it? thank you in advance.

[@BurkeyAcademy]

I don't know how one would do this with GeoDa or without it. What are you thinking of doing? Ranking countries based on GDPPC and calling people neighbors if they are adjacent on the number line? If so, this eems easy to do manually. And then what? It is extremely important to remember that whatever linear modelling you are thinking of doing (correlation, ANOVA, regression) that your definition of neighbors must be exogenous. It is hard to imagine what GDPPC would be exogenous to.

[@josepmarti2025]

In my case I have develop a database where I have the number of Foreign Affiliates that are located in each city (around 1500) in Brazil. I have also the characteristics of each city (salaries, population, education, etc...). So I would like to analyze if there is any kind of spatial correlation about the FDI in each city and the cities characteristics, and therefore observe the effects on its neighbors (direct and indirect). Following your videos, I didn't have problems developing the W matrix in geographic distance terms, using longitud and latitud, and then estimate an OLS reg. However, I would like to know if the economic distance is more relevant than geographic one, for expamle as I said in GDPPC terms. I mean, in economic terms Rio de Jaineiro and Sao Paulo, which are separated for 222 miles aprox., are more similar/neighbors than Rio de Janeiro and the cities that are located close to Rio (in economic terms). So developing this W, I would like to observe if a change in salaries in Rio, will affect the investments in Rio and Sao Paulo if they were neighbors in economic terms. So the question is if there is any option using GeoDa or another software to develop W using a non geographic distance measure.

[@fenggong8299]

@@josepmarti2025 , I'm also working on using non-spatial weights matrix. Have you found a way to develop a non-spatial weights matrix?

[@carolmao5073]

Hi, there. Thanks for your excellent video and it is quite useful for me. By the way, I met something weird words like"1.#INF" when I opened the gwt file (generated by the weights matrices). I would be grateful if you could help me deal with that.

[@BurkeyAcademy]

I haven't seen this before. If you want to email me your data and the settings you used to generate your GWT file, I can probably figure out what is happening. You can find email addresses for me at www.burkeyacademy.com

[@renansereninis]

Link to download the dataset is not working

[@BurkeyAcademy]

Thanks for letting me know-- Google destroyed all of my links, and am trying to fix them all over time. Should be corrected now.

[@jaychang6930]

same problem with clicking "create"

[@BurkeyAcademy]

Thanks for letting me know- I guess it is a bug.

[@avnipravin1936]

Yup, I had the same experience