Chapter 2 Data Collection

1
Chapter 2 - Data Collection

• 2.1 General Principles
• The performance of measuring equipment must be
  known to be adequate.
• Equipment should be recalibrated and the
  precision of measurement devices should be
  monitored.
• The monitoring of this precision is sometimes
  called a gauge R and R study
• Rs stand for Repeatability (variation observed
  when a single operator uses the gauge to measure
  and re-measure one item) and Reproducibility
  (variation is measurement attributable to
  differences among operators)

2
Sampling and Recording

• How much data?
• The question of how much data is needed
  typically depends on the situation. The more
  variation in responses that we expect, the more
  data we need.
• Who should collect the data?
• Those collecting data should be well trained and
  feel that the data they collect will be used in a
  way that is in their best interests. If those
  collecting or releasing data believe it will be
  used against them, it is unrealistic to expect
  them to produce useful data.

3
Sampling and Recording

• Personal biases must not be allowed to enter into
  the data collection process. Sometimes hoped-for
  conditions are given preference over others. A
  solution is to make measurements blind (without
  personnel knowing what set of conditions led to
  an item being measured.)
• In recording data it is important to carry enough
  documentation so that the important circumstances
  surrounding the study could be reconstructed.
  One should not only keep track of the
  experimental variables but of others that may
  later be of interest in the study.

4
Further Note on Bias

• Personal bias may not only come into play in data
  collection, but in data analysis and reporting of
  results as well. For this reason we should
  exercise caution when believing statistical
  results. A good report involves ALL of the
  details concerning data collection and analysis
  to let the reader know that the study was done
  correctly and without bias.
• Example A study shows that areas where country
  music is the most popular also have the highest
  suicide rates.
• Country music also happens to be the most popular
  in areas with an above average rate of poverty.

5
2.2 Sampling in Enumerative Studies

• Simple Random Sample - sample collected from the
  population in such a manner that every collection
  of n items in the populations is equally likely
  to compose the sample.
• Randomness ensures correct statistical results
• Avoids human bias

6
Selecting a Random Sample

• Mechanical methods
• Rely on symmetry and thorough mixing in a
  physical randomizing device.
• The simplest example is drawing slips of paper
  out of a hat. All pieces of paper must be the
  same size/shape and must be well mixed in order
  for it to be truly random.
• Methods using Random Digits
• Use of random number generators built into
  computers. When computers were not widely
  available it was common to use printed random
  digit tables. (Table B.1)

7
Random Number Table (p.785-786)

• Sample across rows or columns without replacement
• each item can only be in the sample once.

8
Example 2.1

• Population of N520 objects are numbered 1
  through 520.
• Collect a random sample of size n8.

9
Example 2.2

• Collect a random sample of size n4 from a
  population of size N6990.

10
2.3 Effective Experimentation

• Response variable
• One that is monitored as characterizing system
  performance (what is measured at the end).
• Supervised variable
• One over which an investigator exercises power,
  choosing the setting(s) for use (what is actively
  manipulated).
• Controlled - a supervised variable that is held
  constant (only has 1 setting)
• Experimental - a supervised variable with
  different settings in a study

11
Definitions (continued)

• Concomitant (or Accompanying) Variable
• observed in an experiment, but neither a primary
  response variable nor a supervised variable.
• Example In studying the fracture strength of
  bricks, you use strength as the response, control
  humidity, use heat as an experimental variable,
  and observe the mold from which each brick came
  (concomitant).
• Extraneous Variable
• Variables in an experiment that affect the
  response and are not of interest to the
  experimenter.

12
Handling Extraneous Variables

• Often the best way to handle extraneous variables
  treat them as controlled variables and hold
  them constant
• Use blocking
• A block of experimental units is a homogenous
  group within which different levels of primary
  experimental variables can be applied and
  compared in a relatively uniform environment.
• Example Put bricks into blocks according to the
  mold they came from. Study only within the
  blocks.

13
Handling Extraneous Variables

• Randomization
• Often means that experimental objects are divided
  up between the experimental conditions at random,
  or that the order of testing is randomly
  determined.
• Tends to spread the effects of uncontrolled
  outside variables evenly across the treatment
  groups.
• Example
• Randomly assign bricks to the heat settings.
  Randomly assign the order in which the heat
  settings are run. Do not apply the same heat to
  all the bricks from one mold.

14
Example 2.3

• 3 molds (blocks)
• 4 temperatures (randomly assigned)

Mold 1
Mold 2
Mold 3
15
2.4 Experimental Plans

• Completely Randomized Experiment - all
  experimental variables are of primary interest
  (none are used for just blocking) and
  randomization is used at every possible point of
  choosing the experimental protocol.
• Units are allocated at random to treatment
  combinations, experimental runs are done in a
  random order, and measuring is done in a random
  order.

16
Example 2.4

• CRE - Effect of amount of hydrofluoric acid and
  time spent in etching bath of glass rods.
• 18 rods (assumed identical)

100 HF
75 HF
50 HF
30 sec
60 sec
120 sec
17
Experimental Plans

• Randomized Complete Block Experiment One
  experimental variable is a blocking factor and
  within each block every setting of the primary
  experimental variable appears at least once.
  Randomization is employed at all possible points.
• Incomplete Block Experiment Not every
  combination of primary response variables appears
  in every block.

18
Example 2.5

• RCBE Effect of fertilizer on corn growth.
• 3 types of fertilizer (A, B, C)
• 4 plots of land in different areas

19
Example 2.6

• IBE Effect of fertilizer on corn growth.
• 3 types of fertilizer (A, B, C)
• 4 plots of land in different areas

20
Steps for Preparing to Collect Engineering Data

• Identify problem
• Understand context
• State objective and scope
• Identify response var. and instrumentation
• Identify possible factors
• Decide how to manage factors

• Create data collection protocol and timetable
• Assign supervision
• Identify technicians and provide guidelines
• Prepare forms and equipment
• Dry run analysis on fictitious data
• Prediction of results

21
Basic Experiment Design

• This is the design I want you to use for your
  project.
• Follows three basic experimental principles.
• Control
• Handling extraneous variables by holding them
  constant
• Randomization
• Randomize the order in which the treatments are
  applied
• Randomize the treatments to units
• Replication
• Must have several experimental units in each
  treatment group

22
Example 2.7

• Assume I tested the effect of music on plant life
  (as done in MythBusters).
• First, 15 small soundproof greenhouses were built
  at a location that would receive good sunlight
  evenly across all greenhouses. Each greenhouse
  was built to the same specifications and using
  the same materials. 150 plants of the same type
  were then randomly selected from a nearby store
  and randomly placed into each greenhouse, 10
  plants to a greenhouse.
• There were three treatments tested heavy metal
  music, classical music, and no music. Each
  treatment was randomly assigned to a greenhouse,
  5 greenhouses per treatment.
• Every plant in every greenhouse was then watered
  twice a day and after a month the amount of
  growth of each plant was measured in inches.
• Are the three experimental principles satisfied?

23
Example 2.7

• Control
• 15 small soundproof greenhouses were built at a
  location that would receive good sunlight evenly
  across all greenhouses.
• Each greenhouse was built to the same
  specifications and using the same materials.
• All 150 plants were of the same type.
• Every plant in every greenhouse was then watered
  twice a day.

24
Example 2.7

• Randomization
• 150 plants were then randomly selected from a
  nearby store and randomly placed into each
  greenhouse, 10 plants to a greenhouse.
• Each treatment was randomly assigned to a
  greenhouse.
• Replication
• 5 greenhouses per treatment