Jessica Kitch			A+

   Ashish: 
   - in general well-done and well reported
   - maybe say something about the EDA in the body of the report
   - tech appendix is nicely done

   BJ: 
   - Excellent, rich exec summary, takes advantage of log-OR's found
   - correct cond indep found from loglinear, but not really interpreted
   - nice tough pursuit of logistic regression, exploiting loglinear findings

Jennifer Fridd			A

   Ashish: 
   - Should probably try to use fewer equations/formulae while
     reporting the analysis, and avoid relying on the equations in the
     descriptive text
   - most findings of the log-linear analysis were not reported in the exec
     summary

   BJ:
   - summary nice
   - when you say "did not signif influence response" in a loglinear
     model what you mean is that the counts were not very different at
     different levels
   - interesting explicit exploration of models conditional on urban,
     after discovering that (race,grade)indep(goals,gender)given(urban)

Alan Charness:			B+

   Ashish:
   - maybe should also report log odds from logistic regression in the exec
     summary and conclusion
   - no EDA
   - need more detailed analysis section
   - could describe the graphical representation better

  BJ:
  - good job but not well backed up by technical appx's.  Eg., where's
    * eda
    * model selection
    * etc.

Alexis Lubensky			B+

  Ashish:
  - No abstract/exec summary
  - Analysis description is sketchy
  - at end, qualitatively suggests some conclusions that could have
    been obtained by logistic regression

  BJ:
  - very good as far as it goes, but missing
    * exec summary
    * logistic regression or data summaries to back up concl's at end
  - summary of analyses a bit too technical, esp incl qq plot

Ty Douglas			B+

  Ashish:
  - Part of exec summary reads like analysis
  - analysis description sounds a bit generic - could make it more 
    specific by naming variables involved, etc.
  - don't know where he went wrong but should have had more 2-factor 
    interactions with "urban" in the loglinear model
  - started on logistic models but gave up for unknown reasons

  BJ:
  - good start but got off on wrong foot by treating sampling zeros in
    race x urban as structural zeros.  distorted model

Jian Bo Yu			B+

  Ashish:
  - doesn't mention possibility of logistic regression analyses
  - could try using step

  BJ: 
  - Nice loglinear analysis and EDA
  - didn't fully interpret graphical model at the end
  - no logistic regression analyses

Kary Myers			B+

  Ashish:
  - Exec summary summarizes the procedure used and not the findings
  - Doesn't do any of the loglinear stuff but does a good complete job
    of addressing the problem she states.
  - emphasizes prediction.  does sample-splitting: train/test.
    possible influence of the reference cited

  BJ:
  - very nice logistic analysis, but misses the loglinear analysis
    that might have guided some of this

Christine Hou			B+

  Ashish:
  - Discusses what she'd expect to find in the data in the exec
    summary and doesn't clearly distinguish her actual findings from
    what she expects to find
  - doesn't mention residual analysis
  - qualititative suggestions of what could be addressed with logistic
    regression

  BJ:
  - Nice job of log-linear analysis 
  - draw the graph and interpret conditional independencies
  - logistic regression follow-ups?
  - model fit?

Katy O'Leary			C-

  Ashish:
  - Exec summary is short and not informative
  - Problem description is vague
  - seems to have misunderstood what was required in analysis
    section. Could have cut-and-pasted parts of her tech appx into
    the analysis section
  - omitted analysis item 10 in tech appx
  - spends disproportionate time on histo's and bar charts
  - seems lost in doing the GLM's.  throws in everything
    indiscriminantly. doesn't look for interactions

  BJ:
  - good exploratory work with hist's / bar charts
  - none of the response variables are appropriate for the analyses
    performed---neither loglinear nor logistic.  This is *serious*
  - no assessment of residuals, though they are plotted
  - summaries difficult to understand.  What is a "criterion
    variable"?  Use plain english.

Craig Weiner			B+

  Ashish:
  - 1/2 sentence sketchy description of findings in the exec summary
  - the tech appx is more like what the analysis section ought to have
    been

  BJ:
  - good job with loglinear models
  - didn't do graph of final model or interp cond indep relations
  - no logistic regression followups

Phoebe Papageorgiou		B+

  Ashish:
  - Didn't do any EDA but well done analysis

  BJ:
  - Nice job on loglinear analysis
  - didn't do graph of final model or interp of cond indep relations
  - no logistic regression analysis

Manoo Halati			B+

  Ashish:
  - nice report of analysis, esp organization and tech appx

  BJ:
  - nice job of loglinear analysis
  - got some cond indep relats but didn't do graph of final model
  - no logisitic regression followups

Yoo Jin Chang			B-

  Ashish:
  (didn't see)

  BJ:
  - nice beginning of loglinear analysis (not completed)
  - good eda, resid plotting
  - no writeup
  - no logistic followup

Huadong Wu			A

  Ashish:
  (didn't see)

  BJ:
  - very nice job
  - loglinear great
  - good EDA, resid analysis
  - good logistic followup
  - would be nice to see more interp of logistic part in writeup