cid-harvard/ecomplexity

ecomplexity

by the Growth Lab at Harvard's Center for International Development

This package is part of Harvard Growth Lab’s portfolio of software packages, digital products and interactive data visualizations. To browse our entire portfolio, please visit growthlab.app. To learn more about our research, please visit Harvard Growth Lab’s home page.

Installation

To install the codes in STATA type the following in the command line: net install ecomplexity, from("https://raw.githubusercontent.com/cid-harvard/ecomplexity/beta/") replace force

ecomplexity's Issues

Normalization of ECI / PCI / COG

ecomplexity/ecomplexity.ado

Lines 312 to 315 in 983290d

	sum eci
	replace eci = (eci-r(mean))/r(sd)
	replace pci = (pci-r(mean))/r(sd)
	replace cog = cog/r(sd)

Currently, PCI and COG are being centered and scaled using the mean and standard deviation of ECI values - would each index not use its own mean and sd for normalization?

Ideally, we would have:

 sum eci 
 replace eci = (eci-r(mean))/r(sd) 
 sum pci
 replace pci = (pci-r(mean))/r(sd) 
 sum cog
 replace cog = cog/r(sd) 

Issues executing the ecomplexity codes

Hello,

I'm a PhD student at George Mason University. I'm working on two projects where I'm attempting to apply, for the first time, these economic complexity methods, and I hope to do many more of these projects in the future. I'm trying to replicate these codes on the BACI dataset to learn how to apply the methods in Stata before I attempt to apply them to a different dataset, but I'm having a few issues executing the codes.

First, I have a question about the read_cepii.do codes. When I "Convert to observatory ids" and merge on i over 10 million obs. did not match and when I merge on j over 8 million obs. did not match. When I "merge to create the export_value and import_value files" 38 million matched and over 94 million did not match. Is that normal? If not, what kind of number should I expect and what might I be doing wrong?

Second, on the atlas_variables.do file, when I "Create touse variable for loading the data into MATA" and then "quietly levelsof origin_id if `touse'==1, local(LOCATION)" I keep getting an error saying ==1 is an invalid name. I tried a workaround and just generated a complex_country variable, but when I "mata exp_cp = rowshape(exp_long,Ncx)," I keep getting an r(3300) error saying my argument is out of range. I thought it might be because my Stata/BE 17 version didn't allow for a big enough matsize for the matrix, so I purchased the Stata/MP 17 version, but I still can't execute the rowshape command. Perhaps it is something I'm doing wrong earlier in the process, but I cannot figure it out and I'm fast approaching some deadlines.

I'd very much appreciate any help or guidance anyone can provide to help me solve these issues. Or, if anyone knows where I might be able to get some help with executing these codes, I'd appreciate a point in the right direction.

Thanks,

Chad Smith

EComplexity Research - Problems with the codes.

Hello
My partner and i, are students of the Asuncion Natinal University from Paraguay.
We are doing a reserch abaut de product space of our country, and we are working in our code with matlab. We can´t reproduce the reluts of the CID Atlas.
We ned te proximity matrix to complete our reserch. If you can help us, it would be great.
We are trying to use te code of the ecomplexity. We use the stata and your codes but we have some prblems.
We have problems with the data source of te Readme file.

net install ecomplexity, from("https://raw.githubusercontent.com/cid-harvard/ecomplexity/beta/") force

We need to know the countries and the products what you use for calculations.

Thanks for help us.
Best Regards.

Time Dimension

Hi Sebastian,

I thought I was in the clear, but I'm running into one final barrier. The sample_file_ecomplexity.do commands work for the SITC rev2 doc downloaded from the Dropbox. But when I upload data I downloaded from Dataverse or dataset I'm using for my own project, I keep getting error codes saying:

1999 Not rectangular, might need to specify the time (t) dimension!
2000 Not rectangular, might need to specify the time (t) dimension!
2001 Not rectangular, might need to specify the time (t) dimension!

The time variables in both are numeric, and I've tried changing the formats to byte, long, double, etc. and also tried everything in the time series menus. Any idea what's going wrong or how I can get over this final hump?

Thanks,
Chad