It would be great to use a trained glmpca to infer factors from new data.
Possibly like:
new_factors <- predict(glmobject, data = new_data)
I imagine the inference part would involve fixing the loadings to be constant, while inferring the factors from the new samples, but there might be details I'm missing.