Dear Dr Zhao / @fzhao70 ,
Thank you very much for making the gamlss_python repository publicly available. I truly appreciate the effort behind this work—it is an incredibly valuable contribution to the community.
I have been using the package to analyse our company data, and I am impressed by how closely the results match those from R GAMLSS. In particular, for families such as BCCGo, BCPEo, and BCTo, the estimates are practically identical, which speaks highly of the implementation.
I wanted to ask whether you have any plans to implement pb() and/or pbz() in the near future.
As you are aware, penalised B-splines are currently not available in the package. From my perspective, this would be a highly impactful addition—especially for applications such as growth curve modelling and centile estimation. The inclusion of spline-based smoothers would allow more flexible modelling of non-linear relationships, enabling local fitting with continuity constraints.
At present, one workaround is to rely on global polynomials. However, this approach has important limitations: small changes or outliers in one region can influence the entire fit, and it lacks the ability to adapt to local structure in the data. This makes it less suitable for applications where capturing local variation is critical.
I would be very interested to hear your thoughts on this, and whether spline support is something you are considering. Thank you again for your excellent work and for making it accessible to the wider community.
Kind regards,
Kerenaftali Klein, PhD
VALD
Australia
Dear Dr Zhao / @fzhao70 ,
Thank you very much for making the gamlss_python repository publicly available. I truly appreciate the effort behind this work—it is an incredibly valuable contribution to the community.
I have been using the package to analyse our company data, and I am impressed by how closely the results match those from R GAMLSS. In particular, for families such as BCCGo, BCPEo, and BCTo, the estimates are practically identical, which speaks highly of the implementation.
I wanted to ask whether you have any plans to implement pb() and/or pbz() in the near future.
As you are aware, penalised B-splines are currently not available in the package. From my perspective, this would be a highly impactful addition—especially for applications such as growth curve modelling and centile estimation. The inclusion of spline-based smoothers would allow more flexible modelling of non-linear relationships, enabling local fitting with continuity constraints.
At present, one workaround is to rely on global polynomials. However, this approach has important limitations: small changes or outliers in one region can influence the entire fit, and it lacks the ability to adapt to local structure in the data. This makes it less suitable for applications where capturing local variation is critical.
I would be very interested to hear your thoughts on this, and whether spline support is something you are considering. Thank you again for your excellent work and for making it accessible to the wider community.
Kind regards,
Kerenaftali Klein, PhD
VALD
Australia