Useful file: 
1. "/groups/icecube/leonbozi/work4/InspectDatabase.py" - a small file used to access a SQL database of your choice. Prints columns/tables in database, number of entries, how many of each PID and examines first event in database.

CSV files to use:
1. "/groups/icecube/leonbozi/work3/stopped_muons_final_reco_RD.csv" - contains ~150k real events that come from database "/groups/icecube/leonbozi/work2/data_real/lvl3_dbs_real/last_one_lvl3/data/last_one_lvl3.db" so the event_no's correspond. These events have been through the event filtration process (particle_score>0.9, muon_score>0.9, stopped_muon_score>0.99) and all 150k have been regressed in azimuth, zenith, energy and stopping_position. All 150k also have had features calculated including n_doms, n_strings, event_duration, average_pulses_per_dom etc. and are included in the csv file.

2. "/groups/icecube/leonbozi/work3/stopped_muons_final_reco_MC.csv" - this is the MC version of the above. Containing ~8 million stopped muon candidates, gone through the same process as the real data. All have been regressed and had features calculated. (be careful, there are a handful of noise events in this csv, they can trip up plotting scripts). These events can be found in the "/groups/icecube/leonbozi/work2/data/lvl3_dbs/last_one_lvl3MC/data/last_one_lvl3MC.db" with consistent event_no's.

3. All of the results.csv in the following folder structures: 
"/groups/icecube/leonbozi/work4/systematics_regression/90eff_results/muons_sys_90"
"/groups/icecube/leonbozi/work4/systematics_regression/95eff_results/muons_sys_95"
"/groups/icecube/leonbozi/work4/systematics_regression/100eff_results/muons_sys_100"
"/groups/icecube/leonbozi/work4/systematics_regression/105eff_results/muons_sys_105"
"/groups/icecube/leonbozi/work4/systematics_regression/110eff_results/muons_sys_110"
These have my preliminary work on the systematic sets containing reconstruction of energy, azimuth, zenith, stopping position. Only containing muongun simulation, but lacking statistics. Naming convention chosen for plotting script purposes. I would recommend doing this analysis again, with more events.

Saved models:
Saved Dynedge models, trained on ~5 million muons (both stopped and through) with learning rate scheduler and almost all early stopping conditions met (converged before max number of epochs ~40 epochs)
1. Energy model - "/groups/icecube/leonbozi/work3/reco_energy_MUONS/train_results_final/energy_lvl2_5m_30e_state_dict.pth"
2. Azimuth model (with kappa) - "/groups/icecube/leonbozi/work3/reco_azimuth_MUONS/training_results/azimuth_lvl2_5m_40e_state_dict.pth"
3. Zenith model (with kappa) - "/groups/icecube/leonbozi/work3/reco_zenith_MUONS/training_results/zenith_lvl2_5m_50e_state_dict.pth"
4. Stopping position (x,y and z) - "/groups/icecube/leonbozi/work3/reco_pos_MUONS/training_results/posxyz_lvl2_5m_50e_state_dict.pth"