Programming Assignment 1

CSC 323 - Data Analysis and Statistical Software

Due: 1/28/99.

The data to be analyzed represents timings (milliseconds) for a new image-processing algorithm. The algorithm has been designed to process images for a digital camera. The times provided are for several distinct images of the same size (i.e. 1152x864 pixels) but of varying complexities. Note that the images are from a standard suite of images used to assess the performance of algorithms of this type. Processing is done in real time and so processing speed is of paramount importance.

Your analysis will involve the determination of the average case and worst case performance for images of this size. See additional details below.

  1. Write a SAS program to analyze these data. Your program should accomplish the following:
    1. Access your data from an external file.
    2. Execute the PRINT and UNIVARIATE procedures with appropriate options.
    3. For PROC PRINT, be sure to use a label for your column heading. Use names that are meaningful. You should generate an appropriate title for your output.
      Note: If necessary, see guide 1 and guide 2.
  2. Write a short analysis (no more than one page) of your output. Your analysis should address the following:
    1. Image processing times are usually assumed to be normally distributed. Is the normality assumption reasonable for this data? Justify your answer.
    2. The average case performance of this algorithm.

      3. Note: Assuming that your sample is representative of the population of images, estimate the mean processing time for images of this size. Your discussion should also provide an estimate of the standard deviation of the population and, given the normality assumption, discuss the implication of a standard deviation of this magnitude.
    3. The worst case performance of this algorithm.

      4. Note: The actual algorithm is not available and so, instead of a traditional analytical analysis, a stochastic analysis must be done. To do this estimate the upper bound on performance by estimating the processing time that corresponds to the 95th percentile.