Uploaded image for project: 'Apache MADlib'
  1. Apache MADlib
  2. MADLIB-797

Grace handling for bad conditioned datasets in Logistic Regression

    XMLWordPrintableJSON

Details

    • Bug
    • Status: Open
    • Minor
    • Resolution: Unresolved
    • None
    • None
    • None
    • None

    Description

      select madlib.logregr_train(
                    'madlibtestdata.ris_part_null_grp2',
                    '__madlib_temp_76339515_1390527247_23401318__',
                    'y', 'x', 'z1, z2', 20, 'irls',
                    '1e-05');}}
      

      has two groups that contains model output all none:

      -[ RECORD 1 ]------------+---------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------
      z1                       | f
      z2                       | t
      coef                     | 
      std_err                  | 
      z_stats                  | 
      p_values                 | 
      odds_ratios              | 
      log_likelihood           | 
      condition_no             | 
      num_rows_processed       | 
      num_missing_rows_skipped | 
      -[ RECORD 2 ]------------+---------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------
      z1                       | t
      z2                       | f
      coef                     | {-1.46617691503338e-06,-0.000782527065408987,-5.08358010746725e-05,2.06285463315915e-05,-6.96067962167662e-05,0.00101502531426917,0.000128813747653072,-0.000160807571215449,1.22870159455695e-05}
      std_err                  | {6.59847010479499e-07,0.000363607214131193,1.55005106870846e-05,1.00858361210063e-05,0.000273987550376967,0.000279729825968701,5.79332541407018e-05,5.34544558376694e-05,1.39167401795097e-05}
      z_stats                  | {-2.22199523790815,-2.15212194642168,-3.27962104610076,2.04529858348852,-0.254050945457184,3.62859166252346,2.22348545000119,-3.00830994714061,0.882894685614691}
      p_values                 | {0.0262836274274374,0.031387751060224,0.00103946607612151,0.0408254302866141,0.799456200079316,0.000284971569781096,0.0261830834290307,0.00262705058776368,0.3772931752287}
      odds_ratios              | {0.99999853382416,0.999217779029047,0.999949165491043,1.0000206287591,0.99993039562628,1.0010155406268,1.0001288220445,0.999839205357629,1.00001228709143}
      log_likelihood           | -922.96699521186
      condition_no             | 3749.39410103314
      num_rows_processed       | 1407
      num_missing_rows_skipped | 15762
      -[ RECORD 3 ]------------+---------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------
      z1                       | t
      z2                       | t
      coef                     | 
      std_err                  | 
      z_stats                  | 
      p_values                 | 
      odds_ratios              | 
      log_likelihood           | 
      condition_no             | 
      num_rows_processed       | 
      num_missing_rows_skipped | 
      -[ RECORD 4 ]------------+---------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------
      z1                       | f
      z2                       | f
      coef                     | {-1.05218385306398e-08,1.56034939242666e-06,2.26640252048327e-07,7.24676864258837e-07,-1.08507458410807e-05,1.00542894495511e-06,-2.59742591851642e-05,-8.56652137523584e-06,2.76317813481785e-06}
      std_err                  | {2.024983513834e-08,6.90270539225558e-07,2.25033629618852e-07,3.11735018295396e-07,4.34071699620892e-06,1.14411524575458e-06,1.83258022034138e-05,9.30050396011515e-06,1.57190457706742e-06}
      z_stats                  | {-0.519601194713844,2.26048962509291,1.00713947702926,2.32465658886017,-2.49975887636938,0.878782927406929,-1.4173600095021,-0.921081417950365,1.75785360964652}
      p_values                 | {0.603341565222738,0.0237908796196956,0.313867752663139,0.0200903302897941,0.0124277861746547,0.379518984824587,0.156377698413318,0.357007920972809,0.078772420881967}
      odds_ratios              | {0.999999989478161,1.00000156035061,1.00000022664028,1.00000072467713,0.999989149313028,1.00000100542945,0.999974026078143,0.999991433515317,1.00000276318195}
      log_likelihood           | -1917.64806277796
      condition_no             | 5606.21687138056
      num_rows_processed       | 2788
      num_missing_rows_skipped | 23106
      

      Attachments

        Activity

          People

            riyer Rahul Iyer
            haying Xixuan (Aaron) Feng
            Votes:
            0 Vote for this issue
            Watchers:
            2 Start watching this issue

            Dates

              Created:
              Updated: