PART I — INTRODUCTION TO DATA COLLECTION

1. What is Data?

Data are facts, measurements, observations, responses, or records collected for the
purpose of analysis.

Examples in Computer Engineering include:

 CPU temperature readings

 processor execution time
 network latency
 bandwidth
 packet loss
 memory utilization
 battery life
 power consumption
 sensor readings
 number of system failures
 response time of an application
 results of user satisfaction surveys

For example, suppose a student wants to compare two algorithms.

The student records:

Algorithm Execution Time

Algorithm A 0.82 s
Algorithm A 0.79 s
Algorithm A 0.84 s
Algorithm B 0.61 s
Algorithm B 0.65 s
Algorithm B 0.59 s

These numbers constitute data.

However, data by themselves do not automatically provide useful information.

The engineering data-analysis process can be viewed as:

Problem → Data Collection → Data Preparation → Data Analysis → Interpretation

→ Decision

Therefore, if the data collected are poor, even sophisticated statistical analysis may
produce misleading conclusions.

2. Why Is Data Collection Important?

Data collection is one of the most important stages of engineering data analysis.

Consider the following situation:

An engineer wants to determine which Wi-Fi configuration provides the

fastest Internet connection.

The engineer measures only one computer, at one location, at 2:00 AM, using one
website.

Can the engineer conclude that the configuration is always faster?

No.

The data may not represent normal operating conditions.

The engineer should consider:

 different computers
 different locations
 different times
 different network loads
 repeated measurements
 different users
 different environmental conditions

This illustrates an important principle:

The quality of the conclusion depends heavily on the quality of the data.

3. Data in Computer Engineering

Computer Engineering combines concepts from:

 computer science
 electronics
 mathematics
 statistics
 electrical engineering
 information technology

Consequently, Computer Engineering students encounter many types of data.

Hardware data

Examples:

 voltage
 current
 temperature
 clock frequency
 power consumption
 battery capacity

Software data

Examples:

 execution time
 memory usage
 CPU utilization
 number of errors
 response time

Network data

Examples:

 latency
 throughput
 packet loss
 connection speed
 number of dropped packets
User data

Examples:

 satisfaction rating
 usability rating
 response to a questionnaire
 frequency of application usage

4. Types of Data

A useful classification is qualitative versus quantitative data.

Qualitative Data

Qualitative data describe categories or characteristics.

Examples:

 operating system: Windows, Linux, macOS

 processor brand: Intel, AMD, Apple
 network type: Wi-Fi, Ethernet, 5G
 satisfaction: satisfied, neutral, dissatisfied

These are generally categorical data.

Quantitative Data

Quantitative data are numerical measurements.

Examples:

 CPU temperature = 68°C

 latency = 25 ms
 memory utilization = 73%
 execution time = 1.24 seconds

Quantitative data can be further classified as:

Discrete data

Countable values.

Example:

Number of system failures = 0, 1, 2, 3,...

Continuous data

Measured values that can theoretically take many values.

Example:

CPU temperature = 65.73°C

PART II — METHODS OF DATA COLLECTION

5. Primary and Secondary Data

One of the first distinctions students should understand is between primary data and
secondary data.
Primary Data

Primary data are collected directly by the researcher for a particular study.

Examples:

 conducting a survey
 measuring CPU temperature
 performing an experiment
 interviewing users
 recording network latency

Suppose you build an Arduino-based temperature monitoring system and collect

temperature readings every minute.

Those readings are primary data.

Secondary Data

Secondary data were previously collected by another person or organization and are
being reused.

Examples:

 government databases
 published research
 company records
 publicly available datasets
 historical system logs
 manufacturer specifications

For example, downloading a publicly available dataset containing CPU benchmark

results means you are using secondary data.

6. Major Methods of Data Collection

Important methods include:

1. Observation
2. Measurement
3. Surveys/questionnaires
4. Interviews
5. Experiments
6. Existing records and databases
7. Automated data logging

Let's examine each.

6.1 Observation

Observation involves systematically watching and recording events or behavior.

Example:

A researcher observes how students interact with a laboratory computer system.

The researcher may record:

 time required to complete a task

 number of errors
 number of attempts
 sequence of actions
Observation can be:

Structured

The researcher already has a specific checklist or recording system.

Unstructured

The researcher records observations without a highly rigid structure.

7. Measurement

Measurement involves obtaining numerical values using an instrument, device, software,

or procedure.

Examples:

 multimeter
 oscilloscope
 temperature sensor
 power meter
 network monitoring software
 performance benchmarking software

Suppose an engineer measures processor temperature every 10 seconds.

The data may look like:

Time Temperature
0 s 42.1°C
10 s 45.7°C
20 s 49.3°C
30 s 53.8°C

Important consideration

Measurements should be:

 accurate
 reliable
 consistent
 properly calibrated
 recorded systematically

8. Interviews

An interview involves asking respondents questions directly.

For example:

"What problems do you usually encounter when using the university's

online learning system?"

Interviews can provide detailed information that numerical measurements may not
capture.

However, interviews can be affected by:

 interviewer bias
 respondent bias
 inaccurate recall
  social desirability bias

9. Questionnaires and Surveys

A questionnaire is a structured set of questions.

A survey is a systematic process of collecting information from respondents.

For example:

"How satisfied are you with the speed of the university Wi-Fi?"

Possible responses:

1. Very dissatisfied
2. Dissatisfied
3. Neutral
4. Satisfied
5. Very satisfied

Surveys are especially useful when collecting information from a large number of people.

10. Automated Data Collection

Computer Engineering students should pay particular attention to automated data

collection.

Modern systems can automatically collect massive quantities of data.

Examples:

 IoT sensors
 server logs
 network monitoring systems
 smart devices
 cloud platforms
 database systems
 wearable devices

For example, a server may automatically record:

Time CPU Memory Response Time

08:00 45% 62% 120 ms
08:01 52% 65% 134 ms
08:02 68% 71% 158 ms

This is useful because automated systems can collect data continuously.

However, automated collection does not guarantee good data.

The system may have:

 faulty sensors
 incorrect timestamps
 missing observations
 duplicated records
 incorrect units
 measurement errors

PART III — POPULATION AND SAMPLE

11. Population
The population is the complete group about which the researcher wants to draw a
conclusion.

Example:

All second-year Computer Engineering students in a university.

That is the population if we want to study their programming proficiency.

12. Sample

A sample is a subset of the population.

For example:

80 selected second-year Computer Engineering students.

Studying the entire population may be impractical because of:

 cost
 time
 limited resources
 accessibility

Therefore, researchers frequently collect data from a sample.

13. Why Is Sampling Important?

Suppose there are 5,000 students in a university.

You want to know their satisfaction with campus Wi-Fi.

Interviewing all 5,000 students may require considerable time.

Instead, you may select 300 students.

But the sample must reasonably represent the population.

This leads to an important principle:

A large sample is not automatically a good sample.

A biased sample can produce biased conclusions regardless of sample size.

PART IV — SAMPLING METHODS

14. Probability Sampling

In probability sampling, members of the population have a known probability of being

selected.

Common methods include:

Simple Random Sampling

Every member has an equal chance of selection.

Example:

From 1,000 students, randomly select 100.

Systematic Sampling

Select every kth member.

Example:

Suppose there are 1,000 students and you need 100.

You could select every:

k= 1000 /100​ =10

student after choosing an appropriate starting point.

Stratified Sampling

Divide the population into groups called strata, then sample from each group.

For example:

 1st year
 2nd year
 3rd year
 4th year

If the research concerns all Computer Engineering students, each year level can be
represented.

Cluster Sampling

Divide the population into clusters and randomly select clusters.

For example:

Select several sections rather than selecting individual students throughout the entire
university.

15. Non-Probability Sampling

Not every member has a known probability of selection.

Examples include:

Convenience Sampling

Select people who are easiest to reach.

Example:

Surveying only classmates because they are readily available.

Purposive Sampling

Select individuals because they meet specific criteria.

Voluntary Response Sampling

People choose whether to participate

PART V — PLANNING AND CONDUCTING SURVEYS

16. What Is a Survey?

A survey is a systematic method for collecting information from people using questions.

A good survey does not begin with:

"Let's create a Google Form."

It begins with:

What exactly do we want to find out?

17. Steps in Planning a Survey

A systematic survey generally involves:

Step 1 — Define the research problem

Example:

Determine the level of satisfaction of Computer Engineering students with

the university Wi-Fi service.

Step 2 — Define the objectives

Possible objectives:

1. Determine student satisfaction.

2. Identify common connectivity problems.
3. Determine whether satisfaction differs by year level.
4. Identify factors affecting Wi-Fi satisfaction.

Step 3 — Identify the target population

Example:

All BS Computer Engineering students enrolled during the current

academic year.

Step 4 — Determine the sample

Decide:

 sample size
 sampling method
 inclusion criteria

Step 5 — Develop the questionnaire

Questions should directly support the objectives.

Step 6 — Test the questionnaire

Conduct a pilot test.

Step 7 — Administer the survey

Distribute the questionnaire.

Step 8 — Collect and clean the data

Check for:

 missing responses
 duplicate responses
 inconsistent responses
 invalid values

Step 9 — Analyze the data

Examples:

 frequency
 percentage
 mean
 median
 standard deviation
 graphs
 statistical tests

Step 10 — Interpret and report the findings

18. Designing Good Survey Questions

A good survey question should be:

 clear
 concise
 specific
 unbiased
 relevant

Poor question

"Don't you agree that the university Wi-Fi is extremely slow and
unreliable?"

This is a leading question.

It encourages a particular answer.

Better question

"How would you rate the reliability of the university Wi-Fi?"

Possible responses:

1. Very poor
2. Poor
3. Fair
4. Good
5. Excellent

19. Avoid Double-Barreled Questions

A double-barreled question asks two things at once.

Poor:

"How satisfied are you with the speed and reliability of the Wi-Fi?"

Speed and reliability are different characteristics.

Better:

"How satisfied are you with the speed of the Wi-Fi?"

and

"How satisfied are you with the reliability of the Wi-Fi?"

20. Avoid Ambiguous Questions

Poor:

"How often do you use the Internet?"

What does "often" mean?

Better:

"How many hours per day do you typically use the Internet?"

Possible categories:

 Less than 1 hour

 1–3 hours
 4–6 hours
 7–9 hours
 More than 9 hours

21. Open-Ended and Closed-Ended Questions

Closed-ended

Respondents select from predetermined answers.

Example:

What operating system do you primarily use?

 Windows
 Linux
 macOS
 Other

Open-ended

Respondents provide their own answer.

Example:

What is the biggest problem you experience with the university network?

Open-ended questions provide richer information but are more difficult to analyze.

22. Likert Scale

A commonly used survey scale is the Likert scale.

Example:

"The laboratory computers are adequate for my programming activities."

1 — Strongly Disagree
2 — Disagree
3 — Neutral
4 — Agree
5 — Strongly Agree

Likert-type questions are useful for measuring attitudes and perceptions.

23. Pilot Testing

Before administering the survey to the entire sample, conduct a pilot test.

For example:

 target sample = 300

 pilot sample = 20–30

Ask pilot participants:

 Are the questions clear?

 Are there confusing words?
 Are response choices complete?
 How long does the survey take?
 Are any questions repetitive?

The pilot test helps identify problems before the actual survey.

24. Survey Bias and Errors

Survey results can be affected by different types of bias.

Selection Bias

The selected sample does not adequately represent the population.

Example:

Only surveying students who regularly use the computer laboratory.

Nonresponse Bias

Selected participants do not respond.

Response Bias

Participants provide inaccurate or socially desirable answers.

Question Bias

The wording of the question influences the response.

PART VI — PLANNING AND CONDUCTING EXPERIMENTS

25. What Is an Experiment?

An experiment is a systematic procedure in which researchers deliberately change one

or more conditions and observe the resulting effects.

Example:

A Computer Engineering student wants to determine whether processor temperature

affects program execution time.
The student can deliberately operate the processor at different temperature conditions and
measure execution time.

26. Observational Study vs Experiment

This distinction is extremely important.

Observational Study

The researcher observes what naturally happens.

Example:

Record the CPU temperatures and execution times of computers during

normal operation.

The researcher does not deliberately manipulate temperature.

Experiment

The researcher deliberately changes a variable.

Example:

Run the processor under controlled temperature conditions and measure

execution time.

The key difference is:

Observation = observe

Experiment = manipulate and observe

PART VII — INTRODUCTION TO DESIGN OF EXPERIMENTS

27. What Is Design of Experiments?

Design of Experiments (DOE) is a systematic statistical approach for planning,

conducting, and analyzing experiments.

DOE helps engineers determine:

 which factors affect a response

 how strongly they affect the response
 whether factors interact
 which combination of conditions produces desirable results

Instead of simply collecting data randomly, DOE provides a structured approach.

28. Example: Testing Network Performance

Suppose we want to determine what affects network latency.

Potential factors include:

 Wi-Fi channel
 distance from router
 number of connected devices
 network bandwidth
 time of day

The response variable might be:

 Network latency in milliseconds

The experiment might investigate:

Factor Level 1 Level 2

Distance 5 m 15 m
Devices 5 20
Channel Channel 1 Channel 6

The researcher can systematically test combinations of these conditions.

29. Important DOE Terminology

Students should understand the following terms.

Response Variable

The outcome being measured.

Example:

Network latency.

Factor

An independent variable that may affect the response.

Example:

Distance from router.

Level

A specific value or condition of a factor.

Example:

Distance = 5 meters

Treatment

A particular combination of factor levels.

Experimental Unit

The basic object receiving a treatment.

Example:

A computer system being tested.

Replication

Repeating an experimental condition.

Randomization

Randomly assigning treatments to experimental units.

Blocking
Grouping similar experimental units to reduce unwanted variation.

30. Example of Factors and Levels

Suppose we are testing laptop battery life.

Potential factor:

Screen brightness

Levels:

 25%
 50%
 75%
 100%

Response:

Battery life in hours.

Another factor could be:

Processor workload

Levels:

 Low
 Medium
 High

Now we have two factors:

1. Screen brightness
2. Processor workload

31. The Three Fundamental Principles of DOE

Three fundamental principles are:

1. Randomization

Randomization helps prevent systematic bias.

Suppose we test two algorithms using 20 computers.

If Algorithm A is always tested in the morning and Algorithm B in the afternoon, time-
of-day effects could influence the results.

Instead, randomly assign testing order.

2. Replication

Replication means repeating an experimental condition.

Suppose Algorithm A is tested only once:

1.25 seconds

We cannot determine whether that represents typical performance.

Instead, run it several times:

 1.25, 1.18, 1.22, 1.27, 1.20 seconds

Replication helps estimate experimental variability.

3. Blocking

Blocking groups experimental units that are similar.

Suppose we compare software performance using:

 high-end computers
 mid-range computers
 low-end computers

Computer performance varies greatly.

We could treat computer type as a blocking variable.

This helps separate the effect of the software from the effect of hardware.

32. Controlled Variables

A controlled variable is a variable kept constant during the experiment.

For example, when comparing two algorithms, we might keep constant:

 input dataset
 computer hardware
 operating system
 compiler
 compiler settings
 background applications

Why?

Because we want differences in execution time to be attributable primarily to the

algorithms.

33. Experimental Error

Experimental measurements naturally contain variation.

For example, repeated execution of the same program may produce:

1.02 s
1.05 s
1.01 s
1.08 s
1.04 s

Why are they not identical?

Possible reasons:

 background processes
 operating-system scheduling
 cache behavior
 temperature
 measurement limitations
 hardware variability

Therefore:
 Variation does not necessarily mean that something is wrong.

A major objective of statistics is to understand and quantify variation.

34. One-Factor Experiment

Suppose we want to determine whether CPU cooling affects processor temperature.

Factor:

Cooling method

Levels:

1. Standard fan
2. Improved fan
3. Liquid cooling

Response:

CPU temperature

We could conduct repeated trials for each cooling method.

Example:

Cooling Method Trial 1 Trial 2 Trial 3

Standard 72 74 73
Improved 67 66 68
Liquid 59 60 58

The next question becomes:

Are the observed differences large enough to conclude that the cooling
methods really differ?

This leads to statistical inference and, later in the course, methods such as analysis of
variance (ANOVA).

35. Two-Factor Experiment

Now suppose we investigate both:

 cooling method
 processor workload

Factors:

Factor A: Cooling

 Standard
 Improved

Factor B: Workload

 Low
 High

This creates four combinations:

 Cooling Workload
Standard Low
Standard High
Improved Low
Improved High

The advantage of a factorial design is that it can investigate not only the individual
effects of the factors but also whether they interact.

36. Interaction

An interaction occurs when the effect of one factor depends on the level of another factor.

Consider:

Does increasing RAM improve program performance?

The answer might depend on the processor.

For an older processor, increasing RAM from 8 GB to 16 GB may have little effect.

For a newer processor running memory-intensive applications, the improvement might be

substantial.

Therefore:

The effect of RAM depends on processor type.

This is an interaction effect.