File: Open Data File: house.txt
Basic: Bivariate: Price as "Y", Assess and Months as "X"
 Hot Spot - Fit Line
  Note Assessed price is significant, but months on market is not
  R-squares are .925 and .064

Model: Fit Model: Price as "Y", Add Assess and Months in "Construct
      Model Effects"
  Fitted regression equation is -44.99 + 1.75*Assess + 0.37*Months
  R-square is .943, adjusted R-square is .938
  overall model is significant (F-test, under "Analysis of Variance")
  both Assess and Months are significant (individual t-tests)
  residual plot?  
   - look for trends, don't make decisions based on possible outliers
  predicted sale price for an assessed value of 80 on the market for 6 months:
      -44.99 + 1.75*80 + 0.37*6 = 97.23


File: Open Data File: speed.txt
Basic: Bivariate: MPG as "Y", MPH as "X"
 Hot Spot - Fit Line
  terrible fit, low R-square, not statistically significant
  Hot Spot - Plot Residuals
    clearly non-linear
Model: Fit Model:  MPG as "Y", Contruct Model Effects: MPH "Macro"
     "Polynomial to Degree" (default degree is 2, a quadratic)
  fit clearly better
  R-square is .919, Adjusted R-square is .912
  F-test of overall model shows significance
  both terms are significant by t-tests
  residual plot?
 try "Polynomial to Degree" with Degree=4
  now R-square is .977, Adjusted R-square is .973
  overall model is significant
  leading term is significant (keep lower order terms for polynomials)


Recordings of levels of a photochemical pollutant (some kind of oxidant) 
made hourly at several recording stations in LA (by the LA Pollution Control 
District). Interest lies in the relationship of pollutant levels to one or 
more meteorological variables (This is a small sample from a very large 
data set). Data recorded are daily maximum levels of the oxidant over a 
series of days, together with morning averages of wind speed, temperature, 
humidity and insolation (a measure of the amount of sunlight) during one 
summer in the early 1990s.

File: Open Data File: oxide.txt
Multivariate: Multivariate: all as "Y"  
  makes a scatterplot matrix and computes correlations
  Which variables look most helpful in predicting the oxidant level?
  Are any of the "independent" variables highly correlated?
  Need to do model building
Mode: Fit Model: Oxidant.Level as "Y", others as effects
  R-square is .798, Adjusted R-square is .756
  overall model is significant
  but not all slopes are significant
    day is least significant, so try again without it
  R-square is .798, Adjusted R-square is .756
    humidity and insolation still not significant (although humidity's
       p-value is much smaller, since it was correlated with day)
    remove insolation (largest p-value) and try again
  R-square is .796, Adjusted R-square is .773 (higher)
    humidity still not significant, so now remove it and try again
  R-square is .777, Adjusted R-square is .761
    overall model is significant
    all effects are significant
    Adjusted R-square is slightly lower, but it is better to have 
      all effect significant
  residual plot looks fine
  regression equation:  oxidant-hat = -5.2 - 0.43*wind + 0.52*temp
  predicted oxidant level for wind speed of 50mph and temp of 80 degrees:
       -5.2 - .43*50 + .52*80 = 14.9