Closed mfarthin closed 6 years ago
It looks like this is the command that is failing Dvec[:Psi_U_i.shape[1]].reshape(Psi_UTv.shape)
. during the call S = np.dot(HX.T, invPsi(HX))
Is the goal to replicate Dvec[:Psi_U_i.shape[1]] into an array of p columns?
@jonghyunharrylee I pushed up a branch mwf/frf_uas_example
where I tried a quick fix. The code runs at least and the linear trend seems to help
@mfarthin Thanks for reporting the error! The latest update to support heterogeneous data sets (by allowing measurement error matrix R as a vector) caused this error. I fixed this based on your implementation and now it is working for stwave duck site example.
@mfarthin I also checked frf_uas_20160722.py in mwf/frf_uas_example worked well. Note that when testing linear/quadratic trend (drift) prior, usually I would set prior variance small compared to a case with no prior.
@jonghyunharrylee when I try to specify a linear trend using the
'drift':'linear'
flag inexample_inv_stwave.py
I get the following tracebackI get a similar complaint execept the size is (69,2) when I try to manually set X. Do I need to set another flag? Thanks