New features in PyDDM

PyDDM has changed a lot since our 2020 eLife paper. Here are some of the new features:

Models are easier to define

Previously, defining a model required building custom classes or using one of many special classes built in to PyDDM. Now, you only need to define a function with simple, consistent syntax.

For example, here is the new way to define a model where drift rate is the product of evidence strength and a fittable parameter:

import pyddm
model = pyddm.gddm(drift=lambda mu,evidence : mu*evidence,
                   conditions=["evidence"],
                   parameters={"mu": (-4, 4)})

Compared to the old way:

import pyddm
class DriftEvidence(pyddm.models.Drift):
    name = "Drift depends linearly on evidence"
    required_parameters = ["mu"]
    required_conditions = ["evidence"]
    def get_drift(self, conditions, **kwargs):
        return self.mu * conditions['evidence']
model = pyddm.Model(drift=DriftEvidence(mu=pyddm.Fittable(minval=-4, maxval=4)),
                    overlay=pyddm.OverlayUniformMixture(umixturecoef=.02))

Learn more about defining models with the gddm function

Simulations are faster

Most simulations should be 10-100x faster than we reported in the eLife paper. Some simulations will be up to 1000x faster, since PyDDM now autodetects more situations where it can use the analytical solver. (And the analytical solver itself is also faster!)

Stimulus coding and accuracy coding

Previously, the top bound in PyDDM always meant “correct” and the bottom bound always meant “error”. This so-called “accuracy coding” was baked into PyDDM. Now, bounds can represent anything you like. For instance, for the upper bound to be a left choice and the bottom bound to be a right choice:

sample = pyddm.Sample.from_pandas_dataframe(..., choice_names=("left", "right"))
model = pyddm.gddm(..., choice_names=("left", "right"))

Learn more about stimulus and accuracy coding

Expanded capabilities of the Model GUI

Previously, the model GUI could be used to plot just the RT distribution. Now, it can also also plot the psychometric and chronometric functions, as well as the bound shape. It can also be expanded to other functions as well. Additionally, the model GUI is now compatible with Jupyter notebooks and Google Colab.

Learn more about the model GUI

Better error messages

When things go wrong, PyDDM is better able to tell you what you need to fix.