Some suggestions of analyzing Boston Housing data

1.  view scatterplots

     (scatterplot-matrix    )

2.    regression
      (regression-model    )

3.   sir

      (sir-i-ii-model     )
 

4. work on the low crime rate group .

++++++++++++++++++++++++++++++++
; finished loading "data-boston.lsp"
> (variables)
(DATA-B NAME-DATA-B)

> name-data-b
("log median house price" " crime rate" "proportion of zoned land" "proportion of nonretail business acres" "charles river, 0,or 1, valued" " nitrogen oxide concentration" "average number of rooms" "proportion of old buildings" "weighted distances to 5 employment centers" "index of accessibility to highways" "property tax rate" "pupil-teacher ratio" "balck proportion of population" "proportion of the poor" "census tract name")

> (def y (nth 0 data-b))
Y
> (def x(select  data-b (iseq 1 13)))
X
> (length x)
14
> (length (nth 0 x))
506
> (scatterplot-matrix data-b)
Error: Too many variables for scatmat   (so I have to select the first 10 variables to plot)
Happened in: #<Subr: #20c4728>
> (scatterplot-matrix x)
Error: Too many variables for scatmat
Happened in: #<Subr: #20c4728>
> (scatterplot-matrix (select data-b (iseq 0 9)))
#<Object: 24daa98, prototype = SCATMAT-PROTO>
 

> (def out-reg (regression-model x Y))

Least Squares Estimates:

Constant                   11.0473      (0.202065)
Variable 0                -1.068486E-2  (1.286071E-3)
Variable 1                 9.891505E-4  (5.177969E-4)
Variable 2                 3.001842E-4  (2.421186E-3)
Variable 3                0.104434      (3.388521E-2)
Variable 4                -9.43747      (1.55374)
Variable 5                 9.962087E-2  (1.623924E-2)
Variable 6                -3.124206E-4  (5.286174E-4)
Variable 7               -0.271113      (3.448673E-2)
Variable 8                0.105369      (1.961935E-2)
Variable 9                -4.730130E-4  (1.256371E-4)
Variable 10               -3.324492E-2  (5.069619E-3)
Variable 11               0.363184      (0.105657)
Variable 12               -2.90076      (0.199475)

R Squared:                0.796614
Sigma hat:                0.186762
Number of cases:               506
Degrees of freedom:            492

OUT-REG

> (send out-reg :plot-residuals)
#<Object: 257c478, prototype = SCATTERPLOT-PROTO>
;; look ok to fit a linear model ???
 

;; Now try sliced inverse regression

> (def out(sir-i-ii-model x y))