Nathaniel Forde
@nathanielforde.bsky.social
The talk ends with a reflection on the nature of craft in statistical modelling, making an an analogy with the practice of writing. Refine and Iteration are key for justifiable and robust conclusions.
November 16, 2025 at 9:09 AM
The talk ends with a reflection on the nature of craft in statistical modelling, making an an analogy with the practice of writing. Refine and Iteration are key for justifiable and robust conclusions.
You can imagine how the latent relationships changes between these states shift as we change the configuration of the agent, and SEMs provide a nice quantified lens on what the implications are for these shifts! How shifted latent states impact downstream choice outcomes!
November 16, 2025 at 9:09 AM
You can imagine how the latent relationships changes between these states shift as we change the configuration of the agent, and SEMs provide a nice quantified lens on what the implications are for these shifts! How shifted latent states impact downstream choice outcomes!
This isn’t only useful for prediction.
It’s increasingly important diagnostically, as we try to understand differences in affect, preference, and function across a growing variety of artificial agents. The post describes how we can impose a hierarchicals structure over the SEM relationships.
It’s increasingly important diagnostically, as we try to understand differences in affect, preference, and function across a growing variety of artificial agents. The post describes how we can impose a hierarchicals structure over the SEM relationships.
November 16, 2025 at 9:09 AM
This isn’t only useful for prediction.
It’s increasingly important diagnostically, as we try to understand differences in affect, preference, and function across a growing variety of artificial agents. The post describes how we can impose a hierarchicals structure over the SEM relationships.
It’s increasingly important diagnostically, as we try to understand differences in affect, preference, and function across a growing variety of artificial agents. The post describes how we can impose a hierarchicals structure over the SEM relationships.
A key idea:
SEMs provide a principled way to abstract over noisy indicators (survey items, behavioural logs, choices, chat responses…) and infer latent constructs representing how an agent perceives or evaluates the world.
SEMs provide a principled way to abstract over noisy indicators (survey items, behavioural logs, choices, chat responses…) and infer latent constructs representing how an agent perceives or evaluates the world.
November 16, 2025 at 9:09 AM
A key idea:
SEMs provide a principled way to abstract over noisy indicators (survey items, behavioural logs, choices, chat responses…) and infer latent constructs representing how an agent perceives or evaluates the world.
SEMs provide a principled way to abstract over noisy indicators (survey items, behavioural logs, choices, chat responses…) and infer latent constructs representing how an agent perceives or evaluates the world.
The conference was energising — kicked off by @inesmontani.bsky.social , with great talks from Pietro Mascolo and others. Also surreal (in a good way) to be back on the UCD campus after so many years.
November 16, 2025 at 9:09 AM
The conference was energising — kicked off by @inesmontani.bsky.social , with great talks from Pietro Mascolo and others. Also surreal (in a good way) to be back on the UCD campus after so many years.
New Blog Post: VAEs vs Bayesian Multivariate models – modelling the complexity of job satisfaction.
nathanielf.github.io/posts/post-w...
It’s not one score. It’s a bundle: how we feel ❤️, how we work 💼, how we think 🧠.
Here’s what happened when we tried to compress that complexity… 🧵
nathanielf.github.io/posts/post-w...
It’s not one score. It’s a bundle: how we feel ❤️, how we work 💼, how we think 🧠.
Here’s what happened when we tried to compress that complexity… 🧵
August 10, 2025 at 5:05 PM
New Blog Post: VAEs vs Bayesian Multivariate models – modelling the complexity of job satisfaction.
nathanielf.github.io/posts/post-w...
It’s not one score. It’s a bundle: how we feel ❤️, how we work 💼, how we think 🧠.
Here’s what happened when we tried to compress that complexity… 🧵
nathanielf.github.io/posts/post-w...
It’s not one score. It’s a bundle: how we feel ❤️, how we work 💼, how we think 🧠.
Here’s what happened when we tried to compress that complexity… 🧵
counterfactual settings to derive different causal estimands. I think this is a powerful combination. If you'll forgive the self-promotion, this is a talk on the subject: nathanielf.github.io/talks/ukrain...
July 24, 2025 at 5:47 AM
counterfactual settings to derive different causal estimands. I think this is a powerful combination. If you'll forgive the self-promotion, this is a talk on the subject: nathanielf.github.io/talks/ukrain...
I like that "measurement error" is baked into the building of a SEM model and that it takes hard to measure concepts seriously. That said i'd agree with other commenters that it's not a tool for causal discovery, but better used to approximate a DAG you believe plausible. I like this framing
July 24, 2025 at 5:47 AM
I like that "measurement error" is baked into the building of a SEM model and that it takes hard to measure concepts seriously. That said i'd agree with other commenters that it's not a tool for causal discovery, but better used to approximate a DAG you believe plausible. I like this framing
Looking forward to @pyconde.bsky.social in Berlin this Year! Will be speaking about @pymc.io -marketing and consumer choice models. Hope to see you there!
berlin.pydata.org/conferences/...
berlin.pydata.org/conferences/...
July 5, 2025 at 1:23 PM
Looking forward to @pyconde.bsky.social in Berlin this Year! Will be speaking about @pymc.io -marketing and consumer choice models. Hope to see you there!
berlin.pydata.org/conferences/...
berlin.pydata.org/conferences/...
🚀 Just added support for multinomial logit & nested logit product choice models to pymc-marketing!
Use a formula API to model product choices & customer preferences — all in a Bayesian workflow 🧠📊
Docs:
🔗 l1nq.com/KRb64
🔗 encr.pw/oa49I
@pymc.io #discrete-choice #bayesian #stats #causal-inference
Use a formula API to model product choices & customer preferences — all in a Bayesian workflow 🧠📊
Docs:
🔗 l1nq.com/KRb64
🔗 encr.pw/oa49I
@pymc.io #discrete-choice #bayesian #stats #causal-inference
June 23, 2025 at 6:25 AM
🚀 Just added support for multinomial logit & nested logit product choice models to pymc-marketing!
Use a formula API to model product choices & customer preferences — all in a Bayesian workflow 🧠📊
Docs:
🔗 l1nq.com/KRb64
🔗 encr.pw/oa49I
@pymc.io #discrete-choice #bayesian #stats #causal-inference
Use a formula API to model product choices & customer preferences — all in a Bayesian workflow 🧠📊
Docs:
🔗 l1nq.com/KRb64
🔗 encr.pw/oa49I
@pymc.io #discrete-choice #bayesian #stats #causal-inference
Python Ireland just released their PyCon 2024 recordings! 📽️
My talk on Causal Inference with @pymc-labs.bsky.social CausalPy package is here. Can we trust individual IV designs. What's the role of CI in industry?
Recording: youtu.be/-C4p4b2cUp8?...
Deck: nathanielf.github.io/talks/pycon_...
My talk on Causal Inference with @pymc-labs.bsky.social CausalPy package is here. Can we trust individual IV designs. What's the role of CI in industry?
Recording: youtu.be/-C4p4b2cUp8?...
Deck: nathanielf.github.io/talks/pycon_...
May 2, 2025 at 5:58 AM
Python Ireland just released their PyCon 2024 recordings! 📽️
My talk on Causal Inference with @pymc-labs.bsky.social CausalPy package is here. Can we trust individual IV designs. What's the role of CI in industry?
Recording: youtu.be/-C4p4b2cUp8?...
Deck: nathanielf.github.io/talks/pycon_...
My talk on Causal Inference with @pymc-labs.bsky.social CausalPy package is here. Can we trust individual IV designs. What's the role of CI in industry?
Recording: youtu.be/-C4p4b2cUp8?...
Deck: nathanielf.github.io/talks/pycon_...
Curious about Multi-level Bayesian Regression Models? Sceptical of Priors? Dubious about of weighted outcomes? Come along to the Python Ireland Meet-up in May, we'll talk survey data, population weights and choosing your own adventure with @pymc.io.
www.meetup.com/pythonirelan...
www.meetup.com/pythonirelan...
April 14, 2025 at 6:40 AM
Curious about Multi-level Bayesian Regression Models? Sceptical of Priors? Dubious about of weighted outcomes? Come along to the Python Ireland Meet-up in May, we'll talk survey data, population weights and choosing your own adventure with @pymc.io.
www.meetup.com/pythonirelan...
www.meetup.com/pythonirelan...
Looking forward to this! Talking about the do-operator and structural causal models in PyMC
January 24, 2025 at 2:36 PM
Looking forward to this! Talking about the do-operator and structural causal models in PyMC
In November I spoke on Bayesian Causal inference and recent developments in the CausalPy eco-system at PyCon Ireland:
nathanielf.github.io/talks/pycon_...
nathanielf.github.io/talks/pycon_...
January 1, 2025 at 3:59 PM
In November I spoke on Bayesian Causal inference and recent developments in the CausalPy eco-system at PyCon Ireland:
nathanielf.github.io/talks/pycon_...
nathanielf.github.io/talks/pycon_...
In September we added examples of structural equation models and confirmatory factor models to the PyMC docs based on a deep-dive blog post of these methods:
Hat-tip @fonnesbeck.bsky.social for the Review
PyMC docs: www.pymc.io/projects/exa...
Deep-dive: nathanielf.github.io/posts/post-w...
Hat-tip @fonnesbeck.bsky.social for the Review
PyMC docs: www.pymc.io/projects/exa...
Deep-dive: nathanielf.github.io/posts/post-w...
January 1, 2025 at 3:59 PM
In September we added examples of structural equation models and confirmatory factor models to the PyMC docs based on a deep-dive blog post of these methods:
Hat-tip @fonnesbeck.bsky.social for the Review
PyMC docs: www.pymc.io/projects/exa...
Deep-dive: nathanielf.github.io/posts/post-w...
Hat-tip @fonnesbeck.bsky.social for the Review
PyMC docs: www.pymc.io/projects/exa...
Deep-dive: nathanielf.github.io/posts/post-w...
In July, jumping off @s3alfisc.bsky.social 's work on pyfixest, we have a deep-dive into @jmwooldridge.bsky.social 's work on Mundlak devices and DiD designs. Highlighting the importance of careful model specification.
nathanielf.github.io/posts/post-w...
nathanielf.github.io/posts/post-w...
January 1, 2025 at 3:59 PM
In July, jumping off @s3alfisc.bsky.social 's work on pyfixest, we have a deep-dive into @jmwooldridge.bsky.social 's work on Mundlak devices and DiD designs. Highlighting the importance of careful model specification.
nathanielf.github.io/posts/post-w...
nathanielf.github.io/posts/post-w...
In June we expanded on the instrumental variable documentation in CausalPy and made use of numpyro samplers to massively speed up sampling times for the IV class in CausalPy:
causalpy.readthedocs.io/en/latest/no...
causalpy.readthedocs.io/en/latest/no...
January 1, 2025 at 3:59 PM
In June we expanded on the instrumental variable documentation in CausalPy and made use of numpyro samplers to massively speed up sampling times for the IV class in CausalPy:
causalpy.readthedocs.io/en/latest/no...
causalpy.readthedocs.io/en/latest/no...
In May I worked with @benvincent.bsky.social to add propensity score weighting methods to CausalPy, jumping off the initial work in Jan/Feb
causalpy.readthedocs.io/en/latest/no...
causalpy.readthedocs.io/en/latest/no...
January 1, 2025 at 3:59 PM
In May I worked with @benvincent.bsky.social to add propensity score weighting methods to CausalPy, jumping off the initial work in Jan/Feb
causalpy.readthedocs.io/en/latest/no...
causalpy.readthedocs.io/en/latest/no...
In April I also posted a deep-dive into the use of spline modelling with @tomicapretto.bsky.social 's Bambi package highlighting the use of hierarchical models over spline components and the trade offs between flexibility and generalisability
nathanielf.github.io/posts/post-w...
nathanielf.github.io/posts/post-w...
January 1, 2025 at 3:59 PM
In April I also posted a deep-dive into the use of spline modelling with @tomicapretto.bsky.social 's Bambi package highlighting the use of hierarchical models over spline components and the trade offs between flexibility and generalisability
nathanielf.github.io/posts/post-w...
nathanielf.github.io/posts/post-w...
In April I spoke at PyData Berlin on Missing Data and Bayesian imputation and met @juanitorduz.bsky.social in real life!
www.youtube.com/watch?v=XUo_...
www.youtube.com/watch?v=XUo_...
January 1, 2025 at 3:59 PM
In April I spoke at PyData Berlin on Missing Data and Bayesian imputation and met @juanitorduz.bsky.social in real life!
www.youtube.com/watch?v=XUo_...
www.youtube.com/watch?v=XUo_...
In Jan/Feb I did a deep-dive into non-parametric causal inference with PyMC, touching on propensity score adjustments and mediation analysis: www.pymc.io/projects/exa...
This was then turned into a modelling webinar on @alex-andorra.bsky.social's podcast : youtu.be/y9BeOr0AETw?...
This was then turned into a modelling webinar on @alex-andorra.bsky.social's podcast : youtu.be/y9BeOr0AETw?...
January 1, 2025 at 3:59 PM
In Jan/Feb I did a deep-dive into non-parametric causal inference with PyMC, touching on propensity score adjustments and mediation analysis: www.pymc.io/projects/exa...
This was then turned into a modelling webinar on @alex-andorra.bsky.social's podcast : youtu.be/y9BeOr0AETw?...
This was then turned into a modelling webinar on @alex-andorra.bsky.social's podcast : youtu.be/y9BeOr0AETw?...
📽 Great fun presenting at Python Ireland PyCON 2024 this weekend! 📽
I spoke about the probing uncertainty in causal estimands using PyMC Labs's and @benvincent.bsky.social 's CausalPy package.
Slides can be found here: nathanielf.github.io/talks/pycon_...
I spoke about the probing uncertainty in causal estimands using PyMC Labs's and @benvincent.bsky.social 's CausalPy package.
Slides can be found here: nathanielf.github.io/talks/pycon_...
November 18, 2024 at 6:43 AM
📽 Great fun presenting at Python Ireland PyCON 2024 this weekend! 📽
I spoke about the probing uncertainty in causal estimands using PyMC Labs's and @benvincent.bsky.social 's CausalPy package.
Slides can be found here: nathanielf.github.io/talks/pycon_...
I spoke about the probing uncertainty in causal estimands using PyMC Labs's and @benvincent.bsky.social 's CausalPy package.
Slides can be found here: nathanielf.github.io/talks/pycon_...
I've not seen that theorem explicitly referenced in model development. It seems right, even just looking at posterior distributions for well specified models seem to take on a normal like distribution for the parameters \theta regardless of prior. More theoretically the Bayesian tends to focus on...
November 8, 2024 at 9:25 PM
I've not seen that theorem explicitly referenced in model development. It seems right, even just looking at posterior distributions for well specified models seem to take on a normal like distribution for the parameters \theta regardless of prior. More theoretically the Bayesian tends to focus on...
Absurd positioning from @irishtimes.bsky.social
October 28, 2024 at 11:16 AM
Absurd positioning from @irishtimes.bsky.social
📉 Interested in Bayesian Psychometrics? 📈 I've added official docs to PyMC on SEM and CFA models here: www.pymc.io/projects/exa...
September 27, 2024 at 6:29 AM
📉 Interested in Bayesian Psychometrics? 📈 I've added official docs to PyMC on SEM and CFA models here: www.pymc.io/projects/exa...