Statistics 601 Fall 2003 : Assignment 1

Q1. Terminate

A company has been forced into across the board layoffs. It has decided to use a summary evaluation criteria made from scores on a number of individual characteristics, such as team work, creativity and productivity. There are 14 such criteria and the best score you can get is a zero, and the worst a 5. An individual's final score on a criteria is the difference between the value assigned and a company provided expectation. For example, if you were a Director, and Directors needed to score 1 on creativity, but you were given a 3, your component score would be plus 2. In other words, exceeding expectations would give you a negative score, but failing to meet them would give you a positive score. The component scores are added together to get a final score. In this context a high score is bad news, you get terminated, and a low score is good news.

The scores for 98 employees are presented in terminate.jmp
Terminate.jmp:

Look at the distribution of scores. Describe in as non-technical way as possible any surprising features in the distribution of the total evaluation score given the fact that each person's data was generated from 14 separate characteristics.

The next four questions are in the back of the case book (pages 229-231). A typical response to a question would involve a couple of short paragraphs with one or two supporting graphics.

A1Q1 :

A1Q3

A1Q4

A1Q5

Q6. Pieces and weight

Part of a company's operations involve moving items around a plant. These items are moved in containers and the plant contains scales to weigh the containers. However, the information reporting system demands "piece counts" for productivity measures, so that these weights have to be multiplied by a density factor, pieces per ounce.

The current density factors may be be inaccurate and management has hired outside consultants to re-evaluate the density factors. There is only one way to do this; sample the items, count them and weigh them.

The consultants developed a methodology that included instructions to data collectors to obtain random samples of approximately 100 items from a large container, and then to accurately weigh these counted items. It takes about 10 minutes to count out roughly 100 items and it would be impracticable to take out more than 200 items because it would slow the plant process down too much.

The data collectors recorded the number of pieces sampled, the weight in pounds of the sample and also the net weight of the container from which the sample was drawn (also in pounds).

The data was recorded by hand on a form that had holders for each digit. It looked something like this:

Data collectors were instructed to fill in all squares because the data was going to be fed through an Optical Character Reader (OCR) machine and then into a database for analysis. For example, entering 132 pieces, 3.17 lbs and a net container weight of 643 lbs would look like this:

Feeding data through the OCR would reduce costs substantially and speed the process. The consultants have come back to you with the initial data. Take a look at it. PW.jmp:

• A. Describe the shape of the distributions and discuss the summary statistics in the context of objectives of the exercise.
• B. Describe any surprising features in the histograms.
• C. Noting the path traversed by the data to make its way into the database, comment on any problems you think may have corrupted the data.
• D. If you have thought of a reason to remove some suspect data, then how sensitive is the pieces per pound estimate to the suspect data?

Last update: Sat Aug 23 12:17:47 EDT 2003

``` Richard_Waterman@AnaBus.com```