Supplemental Material From Lecture 9 (Feb. 2, 1998)

LAST UPDATE: Feb 2, 1998.

  • Fig.1 .. (Postscript File) (PDF File) Is this residual plot good or bad? (Ans.: bad -- there is U-shaped curvature)
  • Fig.2 .. (Postscript File) (PDF File) Is this residual plot good or bad? (Ans.: OK)
  • Fig.3 .. (Postscript File) (PDF File) Is this residual plot good or bad? (Ans.: Not so good -- funnel shaped opening up to the left)
  • Fig.4 .. (Postscript File) (PDF File) Is this residual plot good or bad? (Ans.: Similar to Fig 3, but Prof. Cox says it is an illusion???)
  • Fig.5 .. (Postscript File) (PDF File) Scatterplot of original (x,y) data used to produce Fig. 4. Does this have curvature in it, suggesting that fitting a line is not such a good idea?
  • Fig.6 .. (Postscript File) (PDF File) Same as Fig.5 but with least squared line overlaid. Linear association doesn't seem as bad when the line is put in.
  • Fig.7 .. (Postscript File) (PDF File) Side-by-side boxplots of residuals in Fig.4 when they are broken down by the quintile (fifth) of the x value. The leftmost boxplot is obtained from the residuals corresponding to the 20 smallest x values. The box which is 2nd from left is from the twenty-first through fourtieth x values, etc. Note that the IQR is practically the same for the first 4 boxes, unlike what one would guess from Fig. 4.
  • Fig.8 .. (Postscript File) (PDF File) Can you guess which correlation value goes with which scatterplot?
  • Fig.9 .. (Postscript File) (PDF File) Actual number of boardings (as determined by a monitor) vs. bus driver counts.
  • Fig.10 .. (Postscript File) (PDF File) Deviations from the y=x lines of the data in Fig. 9.

  • Back to Stat 280 Supplemental Material Page

    • Send problems or suggestions to dcox@stat.rice.edu