Linear Regression Project

 
The final report must be no more than 8 pages. SAS code used to
generate the results (plots and tables) must be submitted, organized
into an appendix to the final report. The 8 page limit doesn’t
include the SAS code appendix.  

Discussion is allowed, but you must write the report
independently. Please refer to the Marfan Syndrome Case Study as
a good example of technical writing.
http://www.stat.purdue.edu/~xbw/courses/stat512/notes/marfan.pdf

TA Yucong Zhang, [email protected] will hold office hours on Wed
1-3pm in MATH 541 if you have project related and/or SAS coding
questions. I will answer questions after the class.  

You can find out more about SAS code from SAS support website:
http://support.sas.com/documentation/onlinedoc/base/index.html

***** SAS command to read in data:

data proj;
   infile ‘data-proj.csv’ dlm=”,”;
   input FFMC DMC DC ISI temp RH wind rain area;
run;
proc print data=proj;
run;

***** Dataset Description:

Number of Instances: 517

Missing Attribute Values: None

Attribute information:

  1. FFMC – FFMC index from the FWI system: 18.7 to 96.20
  2. DMC – DMC index from the FWI system: 1.1 to 291.3
  3. DC – DC index from the FWI system: 7.9 to 860.6
  4. ISI – ISI index from the FWI system: 0.0 to 56.10
  5. temp – temperature in Celsius degrees: 2.2 to 33.30
  6. RH – relative humidity in %: 15.0 to 100
  7. wind – wind speed in km/h: 0.40 to 9.40
  8. rain – outside rain in mm/m2 : 0.0 to 6.4
  9. area – the burned area of the forest (in ha): 0.00 to 1090.84

***** Potential Tasks:

1. Response variable is ln(area+1). Create a new column in
the csv file for the response. Save a copy of the csv file with
10 variables. This is the data file actually used for the project.

2. Make scatter plots and have summary statistics using proc
univariate. Examine the dataset (all 10 variables — including the 9
original variables plus the response variable).

3.1. Detect outliers using the measures introduced in
class. Discuss whether you decide to include or exclude certain
observations from the study

3.2. Perform model diagnostics, check residuals. How do you fix the
problems and arrive at your final model?

3.3. Perform variable selection to find the best model

Note task 3.1-3.2-3.3 may need to be run iteratively, and may not
be in this order, to find the best result.

4. Besides write about the process of fitting the regression
model, report your final R^2, AND plot predicted area with the
observed area in the dataset to show how well your model fits the
data. The plot is the one from SAS proc reg, with ln(area+1) as
the response. (You don’t need to make the plot on the original
scale for area).