User:EpochFail/Journal/2011-12-05

Tuesday, Dec. 6th.

I just finished organizing User:Staeiou's dataset and running some logistic regressions.

Scroll to the bottom for my summary

Results

Result's for 420 'vandal' editors

Call:
glm(formula = good_outcome ~ is_anon + personal + nodirectives, 
    data = group_codings)

                 Estimate Std. Error t value Pr(>|t|)    
(Intercept)       0.57159    0.07250   7.884 2.83e-14 ***
is_anonTRUE       0.05404    0.06858   0.788    0.431    
personalTRUE     -0.02676    0.06054  -0.442    0.659    
nodirectivesTRUE -0.06952    0.05942  -1.170    0.243    


Call:
glm(formula = improves ~ is_anon + personal + nodirectives, data = group_codings[group_codings$before_rating <= 
    4, ])
                 Estimate Std. Error t value Pr(>|t|)    
(Intercept)       0.41336    0.09754   4.238 3.26e-05 ***
is_anonTRUE       0.13441    0.08981   1.497    0.136    
personalTRUE     -0.05580    0.08285  -0.674    0.501    
nodirectivesTRUE -0.03947    0.07988  -0.494    0.622    


Call:
glm(formula = contact ~ is_anon + personal + nodirectives, data = group_codings)
                 Estimate Std. Error t value Pr(>|t|)   
(Intercept)       0.10311    0.03194   3.228  0.00134 **
is_anonTRUE      -0.07591    0.03021  -2.512  0.01237 * 
personalTRUE      0.01779    0.02667   0.667  0.50508   
nodirectivesTRUE  0.01673    0.02618   0.639  0.52316   


Call:
glm(formula = good_contact ~ is_anon + personal + nodirectives, 
    data = group_codings)
                 Estimate Std. Error t value Pr(>|t|)
(Intercept)       0.03423    0.02552   1.341    0.181
is_anonTRUE      -0.02205    0.02414  -0.913    0.362
personalTRUE      0.02014    0.02131   0.945    0.345
nodirectivesTRUE  0.02493    0.02092   1.192    0.234


Call:
glm(formula = stay ~ is_anon + personal + nodirectives, data = group_codings)
                  Estimate Std. Error t value Pr(>|t|)    
(Intercept)       0.589011   0.072925   8.077 7.33e-15 ***
is_anonTRUE      -0.049939   0.068982  -0.724    0.470    
personalTRUE      0.006622   0.060896   0.109    0.913    
nodirectivesTRUE  0.051001   0.059776   0.853    0.394    

Result's for 982 'bad faith' editors

Call:
glm(formula = good_outcome ~ is_anon + personal + nodirectives, 
    data = group_codings)
                  Estimate Std. Error t value Pr(>|t|)    
(Intercept)       0.747150   0.053937  13.852   <2e-16 ***
is_anonTRUE      -0.055361   0.052732  -1.050    0.294    
personalTRUE      0.031318   0.035503   0.882    0.378    
nodirectivesTRUE -0.007934   0.035607  -0.223    0.824    


Call:
glm(formula = improves ~ is_anon + personal + nodirectives, data = group_codings[group_codings$before_rating <= 
    4, ])
                 Estimate Std. Error t value Pr(>|t|)    
(Intercept)       0.30488    0.08304   3.671 0.000272 ***
is_anonTRUE      -0.03568    0.08039  -0.444 0.657324    
personalTRUE      0.08101    0.05457   1.484 0.138416    
nodirectivesTRUE  0.01968    0.05381   0.366 0.714758    


Call:
glm(formula = contact ~ is_anon + personal + nodirectives, data = group_codings)
                 Estimate Std. Error t value Pr(>|t|)    
(Intercept)       0.09589    0.02048   4.683 3.23e-06 ***
is_anonTRUE      -0.08619    0.02002  -4.305 1.84e-05 ***
personalTRUE      0.01999    0.01348   1.483   0.1385    
nodirectivesTRUE  0.02532    0.01352   1.873   0.0613 .  


Call:
glm(formula = good_contact ~ is_anon + personal + nodirectives, 
    data = group_codings)
                  Estimate Std. Error t value Pr(>|t|)    
(Intercept)       0.055407   0.015358   3.608 0.000325 ***
is_anonTRUE      -0.048091   0.015015  -3.203 0.001405 ** 
personalTRUE      0.007076   0.010109   0.700 0.484101    
nodirectivesTRUE  0.011429   0.010139   1.127 0.259912    


Call:
glm(formula = stay ~ is_anon + personal + nodirectives, data = group_codings)
                 Estimate Std. Error t value Pr(>|t|)    
(Intercept)      0.427547   0.058728   7.280 6.85e-13 ***
is_anonTRUE      0.001126   0.057417   0.020    0.984    
personalTRUE     0.005804   0.038657   0.150    0.881    
nodirectivesTRUE 0.035127   0.038770   0.906    0.365    

Result's for 702 'test' editors

Call:
glm(formula = good_outcome ~ is_anon + personal + nodirectives, 
    data = group_codings)
                 Estimate Std. Error t value Pr(>|t|)    
(Intercept)       0.32041    0.05007   6.400 2.86e-10 ***
is_anonTRUE      -0.19191    0.04950  -3.877 0.000116 ***
personalTRUE     -0.06622    0.02799  -2.366 0.018269 *  
nodirectivesTRUE -0.02851    0.02860  -0.997 0.319314    


Call:
glm(formula = improves ~ is_anon + personal + nodirectives, data = group_codings[group_codings$before_rating <= 
    4, ])
                 Estimate Std. Error t value Pr(>|t|)    
(Intercept)       0.51628    0.08999   5.737 2.71e-08 ***
is_anonTRUE      -0.20262    0.09047  -2.240   0.0260 *  
personalTRUE     -0.08154    0.06867  -1.187   0.2362    
nodirectivesTRUE -0.13342    0.06321  -2.111   0.0357 *  


Call:
glm(formula = contact ~ is_anon + personal + nodirectives, data = group_codings)
                  Estimate Std. Error t value Pr(>|t|)    
(Intercept)       0.102792   0.029816   3.448 0.000600 ***
is_anonTRUE      -0.092998   0.029477  -3.155 0.001674 ** 
personalTRUE      0.006664   0.016669   0.400 0.689437    
nodirectivesTRUE  0.058796   0.017034   3.452 0.000591 ***


Call:
glm(formula = good_contact ~ is_anon + personal + nodirectives, 
    data = group_codings)
                 Estimate Std. Error t value Pr(>|t|)   
(Intercept)       0.08055    0.02552   3.156  0.00167 **
is_anonTRUE      -0.07780    0.02523  -3.083  0.00213 **
personalTRUE      0.01449    0.01427   1.015  0.31031   
nodirectivesTRUE  0.04365    0.01458   2.993  0.00286 **


Call:
glm(formula = stay ~ is_anon + personal + nodirectives, data = group_codings)
                 Estimate Std. Error t value Pr(>|t|)    
(Intercept)       0.66019    0.07762   8.505  < 2e-16 ***
is_anonTRUE      -0.27420    0.07674  -3.573 0.000377 ***
personalTRUE     -0.11179    0.04339  -2.576 0.010195 *  
nodirectivesTRUE  0.02281    0.04435   0.514 0.607095    

Result's for 347 'good faith' editors

Call:
glm(formula = good_outcome ~ is_anon + personal + nodirectives, 
    data = group_codings)
                 Estimate Std. Error t value Pr(>|t|)    
(Intercept)       0.54997    0.06567   8.375 1.45e-15 ***
is_anonTRUE      -0.22114    0.06548  -3.377 0.000817 ***
personalTRUE      0.02469    0.06455   0.382 0.702373    
nodirectivesTRUE  0.03012    0.06340   0.475 0.635100    


Call:
glm(formula = improves ~ is_anon + personal + nodirectives, data = group_codings[group_codings$before_rating <= 
    4, ])
                 Estimate Std. Error t value Pr(>|t|)    
(Intercept)       0.41501    0.09437   4.398 2.63e-05 ***
is_anonTRUE      -0.09934    0.10140  -0.980    0.330    
personalTRUE     -0.13024    0.11243  -1.158    0.249    
nodirectivesTRUE  0.07892    0.11049   0.714    0.477    


Call:
glm(formula = contact ~ is_anon + personal + nodirectives, data = group_codings)
                  Estimate Std. Error t value Pr(>|t|)    
(Intercept)       0.174159   0.040362   4.315 2.09e-05 ***
is_anonTRUE      -0.131793   0.040247  -3.275  0.00117 ** 
personalTRUE      0.009522   0.039672   0.240  0.81046    
nodirectivesTRUE  0.086722   0.038970   2.225  0.02671 *  


Call:
glm(formula = good_contact ~ is_anon + personal + nodirectives, 
    data = group_codings)
                 Estimate Std. Error t value Pr(>|t|)    
(Intercept)       0.14026    0.03906   3.591 0.000378 ***
is_anonTRUE      -0.08553    0.03895  -2.196 0.028781 *  
personalTRUE     -0.01377    0.03839  -0.359 0.719990    
nodirectivesTRUE  0.07412    0.03771   1.965 0.050198 .  


Call:
glm(formula = stay ~ is_anon + personal + nodirectives, data = group_codings)
                 Estimate Std. Error t value Pr(>|t|)    
(Intercept)       0.70754    0.06670  10.608  < 2e-16 ***
is_anonTRUE      -0.24794    0.06651  -3.728 0.000226 ***
personalTRUE      0.06999    0.06556   1.068 0.286437    
nodirectivesTRUE  0.04712    0.06440   0.732 0.464845    

Summary

• Vandals (blatant)
  • No interesting significant effects here
• Bad faith
  • No directive increases the probability of making contact (marginal significance)
• Testers
  • Personal messages reduce the probability of a good outcome
  • No directives messages reduce the probability that an editor will improve.
  • No directives increases the probability of contact and good contact
  • Personal messages reduce the probability of staying to make more edits
• Good faith
  • No directives increases the probability of contact and good contact.