dfm / tess-atlas

MIT License

9 stars 8 forks source link

How can we improve fits with multi-modal posteriors? #184

Closed avivajpeyi closed 2 years ago

avivajpeyi commented 2 years ago

Eg TOI 237

ExFop page SNR~7.7 is low?

planet_0_posteriors 68747470733a2f2f692e6962622e636f2f43485266527a4a2f53637265656e2d53686f742d323032322d30322d30372d61742d31302d30372d31302d706d2e706e67

avivajpeyi commented 2 years ago

Could it be due to two different chains at different spots?

avivajpeyi commented 2 years ago

Interestingly p and t_max trace plots are also multimodal for both traces (sold and dashed are two different traces)

Could this weirdness just be due to the low SNR? diagnostic_flux_vs_time

avivajpeyi commented 2 years ago

plot other cases
is difference in t0 --> 0.6 similar

avivajpeyi commented 2 years ago

split samples on the difference modes
plot phase plots for different modes

avivajpeyi commented 2 years ago

tweak plot settings till we can figure out what is going on

avivajpeyi commented 2 years ago

Closing for now