- [hddm](https://hddm.readthedocs.io/en/latest/index.html) # Problem ## How to install hddm on macOS (or/and Linux)? ```python # Esin on googlegroup: The steps that worked for me (with Anaconda-Navigator 1.9.7 installed, in case that matters, but all commands executed in the regular Terminal): conda create --name py36 python=3.6 conda activate py36 conda install conda-build conda install pymc conda install pandas patsy export MACOSX_DEPLOYMENT_TARGET=10.11 # exclude this line for Linux pip install cython pip install hddm ``` - [Installation instructions for OSX](https://groups.google.com/g/hddm-users/c/bdQXewfUzLs/m/iVasvmW3BwAJ) - [Installation problem · Issue #41 · hddm-devs/hddm · GitHub](https://github.com/hddm-devs/hddm/issues/41) ## Parameters in HDDM HDDM uses 1 (others usually use 0.1) as the noise coefficient or scaling parameter. Thus, to compare with other papers/models, we need to multiply HDDM's estimates by 0.1. The `z` (bias/starting point) parameter in HDDM is relative to boundary parameter `a` (it ranges from 0 to 1). Ratcliff reports `z` as an absolute value (ranges from 0 to `a`). To transform HDDM `z` to Ratcliff's notation, use `a * z`. - [Parameters](https://hddm.readthedocs.io/en/latest/howto.html?highlight=sample#compare-parameters-to-other-papers) # Use HDDMRegressor to fit between-person effects Standardize or mean-center between-subjects predictors before fitting model. Michael Frank ran simulations to show it's possible with `HDDMRegressor`. - [Using HDDMRegressor to fit between-person effects](https://groups.google.com/g/hddm-users/c/C96uz2_xFl8/m/VgIehFG4AQAJ) ## Error "AttributeError: Deterministic z's value cannot be set." when fitting models to individual subjects Try setting `p_outlier=0.05` to remove outliers. If not, remove `model.find_starting_values()` but make sure to increase the number of samples and specifically the burn in. - [difficulties fitting model to individual participant data](https://groups.google.com/g/hddm-users/c/HsBZDr3_CTQ/m/iwyY8IYrAQAJ) # Resources - [hddm docker image](https://hub.docker.com/r/hcp4715/hddm) - [How-to — HDDM 0.9.2 documentation](https://hddm.readthedocs.io/en/latest/howto.html?highlight=sample#how-to)