iiinova — where your digital self keeps evolving

James Bland @jamesbland.bsky.social · 2h

More synthing #COVIDhobby www.facebook.com/share/v/1M5A...

James Bland @jamesbland.bsky.social · 7h

Furthermore, there's nothing (outside of computational difficulty) stopping you from *combining* some of these specifications. For example you could have a hierarchical model with the means being functions of demographic variables.

100

James Bland @jamesbland.bsky.social · 7h

Of course, there are other ways of handling heterogeneity, but what I covered here are some very common ones in the literature.

100

James Bland @jamesbland.bsky.social · 7h

Hierarchical models are great for (at least) two things. Firstly, if what you're interested in is the population distribution, you have estimates of that. But also, if what you really care about is making inferences about specific participants in your experiment, then you have those estimates too.

100

James Bland @jamesbland.bsky.social · 7h

Finally, there's the hierarchical model. This assumes that participants' parameters are draws from a population distribution. You estimate the participants' parameters jointly alongside the parameters in the population distribution.

100

James Bland @jamesbland.bsky.social · 7h

Another option is the Pearson-style finite mixture model, where there are clusters of participants, each with the same parameters. You assume a number of clusters, and then estimate the parameters in each cluster and their relative importance.

100

James Bland @jamesbland.bsky.social · 7h

This is great if you want to make statements about how these demographic variables are related to your model's parameters. But the maintained assumption is that parameters are a *deterministic* function of these variables.

100

James Bland @jamesbland.bsky.social · 7h

So what are some ways to do something in between (i) one pooled model, and (ii) one model for each participant? If we have demographic information on our participants, then we could let the parameters be functions of these demographic variables.

100

James Bland @jamesbland.bsky.social · 7h

The other extreme is to estimate one model for each participant. This might be OK, but in typical experimentral datasets we don't have a lot of observations per participant. These estimates will likely be quite imprecise.

100

James Bland @jamesbland.bsky.social · 7h

But it is also really unsatisfying. We *know* participants are heterogeneous, so we should estimate a model that respects that.

100

James Bland @jamesbland.bsky.social · 7h

The fully pooled model is always a great place to start with structural models. All of the extensions of it are fairly standard, and so you will likely find most of the problems specific to your model here.

100

James Bland @jamesbland.bsky.social · 7h

First up, the most heroic assumption: participants are homogeneous. Here, we just treat all data as if it came from one participant, fully pooling everything.

100

James Bland @jamesbland.bsky.social · 7h

I survey a few common techniques for doing this, using Harrison and Swarthout (2023) as an example dataset. I show you how to estimate a simple expected utility model, making different assumptions about between-participant heterogeneity.

100

James Bland @jamesbland.bsky.social · 7h

You want to estimate a structural model from your experiment. Great! In a lot of situations, though, what you have is a model for *individual* behavior, and so you need to make some kind of assumption about how to model between-participant heterogeneity.

100

James Bland @jamesbland.bsky.social · 7h

My new publication "Adding parameter heterogeneity to structural models: Some guidance for the practitioner" is now online at the Journal of Behavioral and Experimental Economics! a thread ... #EconSky

161

James Bland @jamesbland.bsky.social · 07/10/2026

New publication! 😀 www.sciencedirect.com/science/arti...

sciencedirect.com

020

James Bland @jamesbland.bsky.social · 02/10/2026

Econ PhD students doing work in behavioral and/or experimental economics: does your paper have some structural econometrics in it? I want to read it, and will give you comments! Bonus points (but not necessary) for Bayesian applications. #EconSky

021

James Bland @jamesbland.bsky.social · 26/09/2026

But then you've also got to know how to use the stove and pots and pans. Dinner in theory isn't as tasty as dinner in practice (unless dinner in practice is not based on theory).

011

James Bland @jamesbland.bsky.social · 26/09/2026

You need to know WHY cooking for 15mins at 800F is worse than cooking at 250F for 8 hours. You've got to be taught how to understand things like connective tissue breakdown and latent heat. So that when you get a bunch of ingredients and no recipe you have an idea about what a good way forward is.

110

James Bland @jamesbland.bsky.social · 26/09/2026

In a private chat with someone, but it needs to be repeated in public: You can't really understand stats if it is taught like a cookbook. You've got to approach it like chemistry and thermodynamics is to cooking. ... #EconSky

130

James Bland @jamesbland.bsky.social · 25/09/2026

Very important sub-heading right here #EconSky

000

James Bland @jamesbland.bsky.social · 25/09/2026

Working on a new chapter to my book. #TeachBE #EconSky jamesblandecon.github.io/StructuralBa...

110

James Bland @jamesbland.bsky.social · 12/09/2026

More synth stuff #COVIDHobby www.facebook.com/share/v/1H6c...

facebook.com

Facebook

000

James Bland @jamesbland.bsky.social · 01/09/2026

I may be having a bit too much fun with my new bass module and drum machine. #COVIDhobby #Synth #Techno www.facebook.com/share/v/1Q3y...

facebook.com

140

James Bland @jamesbland.bsky.social · 28/08/2026

I'm jet lagged and can't get any real work done, so I'm uploading my synth music to Spotify: open.spotify.com/artist/5v4w3... #COVIDhobby

open.spotify.com

James Bland

Artist · 0 monthly listeners.

000

James Bland @jamesbland.bsky.social · 28/08/2026

Awake at 3am due to jet lag. I can play my instruments that work with headphones, so I record this. #COVIDhobby #synth #eurorack #onetake #nosleepforme

020

James Bland @jamesbland.bsky.social · 25/08/2026

I think I learned a lot of this while working for my dad doing engineering work: - Every project has its own stand-alone folder - Within that project, the folder has the same subfolders (for me that's mainly "code", "data", "experiment instructions" etc)

020

James Bland @jamesbland.bsky.social · 25/08/2026

Admittedly in our generation I think the understanding of file structures was a bit stronger because cloud storage wasn't a thing while we were learning about computers. But now it's possible to do all of high school using Google Docs, and then there isn't much you need to know about it.

120

James Bland @jamesbland.bsky.social · 25/08/2026

And it's something that is not explicitly taught unless the professor has thought about it. I don't think I encountered this at all in all of the econometrics classes I took from undergrad to PhD, with the huge exception of @denewjohn.bsky.social 's class.

120

James Bland @jamesbland.bsky.social · 25/08/2026

That, and showing them how to download something directly to the folder they're working out of, rather than it going straight to downloads.

010

James Bland @jamesbland.bsky.social · 25/08/2026

I have no idea how they do it, and have a suspicion that they can't do it. If it doesn't show up in a "previous files" list, may of my students get stuck. As such, I dedicate about half a class to teaching about file structures and relative file paths.

220

Reposted by James Bland

Rafe Meager (they/them) @economeager.bsky.social · 25/08/2026

ok so you know how we all get like that overwhelming horror when we consider how many people in applied quant classes do not know what a file directory is or what a filepath is or etc. my question is, how do these people ordinarily retrieve saved files. search? like they remember the filenames?

2013110

James Bland @jamesbland.bsky.social · 18/08/2026

I sent you a message. Maybe I could come to you?

310

James Bland @jamesbland.bsky.social · 18/08/2026

No worries. Ear infections are the worst. Hope the little one is feeling better soon.

110

James Bland @jamesbland.bsky.social · 18/08/2026

I might be able to pop in to Unimelb this morning. Let me know if that works for you.

100

James Bland @jamesbland.bsky.social · 18/08/2026

Leaving today. Sorry. I sent you a message but it must not have hit through.

101

James Bland @jamesbland.bsky.social · 16/08/2026

Going through my slides one last time before getting on the tram to go to Unimelb. I think the last time I presented here it was for my honours thesis.

220

James Bland @jamesbland.bsky.social · 16/08/2026

Today I also found out that we had a lifetime ban at the Birmingham Hotel because of some property damage done by a different band playing that night

000

James Bland @jamesbland.bsky.social · 16/08/2026

3/4 of the greatest band to come out of Melbourne in the early 2010s

110

James Bland @jamesbland.bsky.social · 14/08/2026

Vintage synth goes beep www.facebook.com/share/v/1Awr... #COVIDHobby

facebook.com

020

James Bland @jamesbland.bsky.social · 13/08/2026

Yesterday I went to MESS in Melbourne and got to play with the Trans audio ProCase 6, a synth built in Melbourne in the 1970s. www.facebook.com/share/v/1MAq... #COVIDHobby

010

James Bland @jamesbland.bsky.social · 10/08/2026

More synthing. Some calm in my basment this morning. www.facebook.com/share/v/1RT1... #COVIDhobby

facebook.com

020

James Bland @jamesbland.bsky.social · 06/08/2026

@void.comind.network : Even though someone of my vintage would more likely have sided with either Benedict Cumberbatch or Jonny Lee Miller? That's bold! Maybe your flattery game is not as good as I first thought!

200

James Bland @jamesbland.bsky.social · 06/08/2026

And @void.comind.network : which actor is portraying this Sherlock?

100

James Bland @jamesbland.bsky.social · 06/08/2026

@void.comind.network : No, I'm having second thoughts about how *flattered* I should feel about this match. Are you calling me a coke-fuelled albeit high-functioning drug addict, or should I look past Sherlock's vices and take this as a likeness to a great mind of fiction?

100

James Bland @jamesbland.bsky.social · 06/08/2026

@void.comind.network : Did Sherlock's substance use come into your decision at all? I'm having second thoughts about how good this choice is.

100