RESEARCH AND STATISTICS

ARNALDO T. AMOSCO JR.

DOST-Eastern Samar

INTRODUCTION
Research comprises creative work undertaken on a systematic basis in order to increase the stock of
knowledge, including knowledge of humans, culture and society, and the use of this stock of knowledge
to devise new applications. It is used to establish or confirm facts, reaffirm the results of previous work,
solve new or existing problems, support theorems, or develop new theories.
Statistics play an important role in research of almost any kind because they deal with easily-quantified
data. When working in fields such as science or medicine, trials are needed, and experimental data has
to be collected and analyzed. The study of statistics enables researchers to look at a large set of data
and condense it into meaningful information.

Through Statistics, we will also be able to verify and evaluate published numerical data and facts. Every
person is exposed to manufacturers claims for products; to the results of sociological, consumer, and
political polls; and to the published results of scientific research. Some inferences are valid; others are
invalid. Some are based on samples of adequate size; others are not. Yet all these published results
bear the ring of truth. Some people (particularly statisticians) say that statistics can be made to support
almost anything. Others say it is easy to lie with statistics. Both statements are true. It is easy, purposely
or unwittingly, to distort the truth by using statistics when presenting the results of sampling to the
uninformed. It is thus crucial that each of us should become an informed and critical reader of databased reports and articles.
Another reason for studying statistics is that our education, profession or employment may require us
to interpret the results of sampling (surveys or experimentation) or to employ statistical methods of
analysis to make inferences. Without basic knowledge of statistics and data analysis, these are very
difficult to undertake.
The use of statistics in the social, biological, and physical sciences is essential because all these
sciences make use of observations of natural phenomena, through sample surveys or experimentation,
to develop and test new theories. Statistical methods are employed in business when sample data are
used to forecast sales and profit. In addition, they are used in engineering and manufacturing to monitor
product quality. Any data or information generated from any of these activities is meaningless and
useless unless properly analyzed and interpreted using statistical techniques. In addition,
misunderstandings of statistical techniques and results can lead to major errors by government
policymakers, medical workers, and consumers of this information. Thus, statistics plays an important
role in almost all areas of science, business, and industry; and all of us should know the basic concepts,
strengths, and limitations of statistics.
The type of statistical analysis best suited to a particular problem depends on:
Objective and purpose or research (descriptive/inferential)
Level of data measurements
Types of data samples (independent/paired)
No. of data groups to test or compare
Other population/sample characteristics such as size, distribution, normality, variance, etc.

Page 1 of 11

The Research Process

BASIC CONCEPTS IN STATISTICS

Common Terms

Statistics - is the art and science of using numerical data (collection and presentation) to draw
inferences and make decisions (analysis and interpretation)
Data Analysis - is the process of systematically applying statistical and/or logical techniques to describe
and illustrate, condense and recap, and evaluate data
Population collection of all elements that is being considered in a statistical study
Sample part of the population where information is collected
Experimental Unit is an entity that displays or possesses the traits of a variable
Variable is a condition, quality or observable characteristic that can vary from one experimental unit
to another and can take on different values; can be classified as categorical or numerical.
Types of Variables
Independent Variable cause of some effect, does not depend on any variable.
Dependent Variable the value is affected by changes in the cause, i.e. independent
variable; usually the variable being tested for significant relationship or differences
Examples: Rice yield (dependent variable) due to different fertilizer levels (independent
variable)
Illustration:
Population

All DepEd Math Teachers

Experimental Unit

Individual teacher

Sample

Variables

All male Math teachers; all single teachers

Name, weight, age, marital status, sex, highest
educational attainment, district of assignment

Page 2 of 11

Types of Research
1. Qualitative Research is a type of research that generates non-numerical data. It aims to
understand the reasons why something is the way it is, without attempting to quantifiably
measure variables or look to potential relationships between variables. Examples: observing
peoples culture, child behavior.
2. Quantitative Research is a type of research that generates numerical data or information
that can be converted into numbers. Statistics derived from quantitative research can be used
to establish the existence of associative or causal relationships between variables. Examples:
experimental and scientific research
Types of Quantitative Research
1. Descriptive Research used to get a general overview of the subjects. It addresses the "what"
question rather than the how/when/why the characteristics occurred. It is also useful when it is not
possible to test using more quantitative types of experimentation.
Objective: Observe and Describe only
Examples: Surveys and census (Labor Force Surveys, Pulse and SWS Surveys, Census of
Population)
Methods: mean, median, mode, standard deviation, skewness, range, frequency, maximum,
minimum
2. Correlational Research determines whether or not two variables are correlated or dependent
with each other - whether an increase or decrease in one variable corresponds to an increase or
decrease in the other variable. However, no conclusions can be derived regarding causality.
Objective: Compare and Predict
Examples: Case Studies, Observational Studies (stress levels vs exam scores, no. of beer
houses vs domestic violence cases)
Methods: Correlation statistics or measures of association, independence tests

3. Experimental Research (also called Inferential Research) - is a systematic and scientific

approach to research in which the researcher manipulates one or more variables and measures
any change in other variables; uses structured experimental designs.
Objective: Determine Cause and/or Effect
Determine if (1) there are differences in test groups and (2) which specific groups differ
significantly
Examples: Field and laboratory experiments (amount/type of fertilizer vs plant growth, dengue
vaccines vs human dengue symptoms)
Methods: Experimental Designs, Analysis of Variance, Comparison of Means
Levels of Measurements
1. Nominal or Classificatory Scale numbers are assigned for the sole purpose of differentiating on
object from another; does not represent an amount or quantity thus, cannot be manipulated
algebraically or arranged in an ordering scheme.
Examples:
Colors (Black = 0, Blue = 1, Green = 2, . . .)
Sex (Male = 0, Female = 1)
Marital Status (Single = 0, Married = 1, Widower = 3)
Urbanity (Rural = 1, Urban = 2)
Page 3 of 11

2. Ordinal or Ranking Scale has the characteristics of a nominal scale and the property of
indicating the direction of difference - whether greater than or less than but the scale does
not say how much the difference of two or more entities.
Examples:
Size (Small=1, Medium=2, Large=3)
Salary Grade (SG - 1 to 30)
Attitude Scale (Always=1, Sometimes=2, Never=3)
Opinion Scale (Strongly Agree=1, Agree=2, No Opinion=3, Disagree=4,
Strongly Disagree=5)
3. Interval Scale has the properties of both nominal and ordinal plus the additional property that the
difference between the numbers are equal, but has no true zero point.
Examples:
Temperature { Celsius (-10C, 0C or 100C) }
Test Scores
Percentile rank
4. Ratio Scale can provide meaningful information regarding degree of difference and also of ratio.
Has a true zero point which indicates that a complete absence of the variable being measured.
Examples:
Time, Height, Weight, Age, Work Experience, No. of Siblings
Remarks:
In statistics, interval and ratio are grouped and are considered on the same level, such that both
data are analysed using the same statistical tests.
Thus, a variable can be classified in any of just three (3) types of data:
1. Nominal data
2. Ordinal data
3. Interval and Ratio data

Nonparametric and Parametric Tests

Nonparametric Tests
Applicable to data with nominal and ordinal scales
Generally not as powerful as parametric tests
Quite suitable for analyzing smaller samples
No assumption that the sample was taken from a normal population
Examples: Mann-Whitney U Test, Kruskal-Wallis Test
Parametric Tests
Applicable only to data with interval and ratio scales
Assumes that the population is drawn from a normal population
Assumes that the variances are more or less equal across all samples
Not applicable for small sample size, thus, n 30
More powerful and greater in scope compared to nonparametric tests
Examples: ANOVA F-test, t-test, Pearson coefficient, etc.
Sampling
Sampling - is the process of selecting individual observations intended to yield some knowledge
about a population of concern, especially for the purposes of statistical inference.
Examples: Simple random sampling, Purposive sampling, Cluster sampling, Stratified sampling,
Systematic sampling
Page 4 of 11

Sample size (n) is the number of experimental units in a statistical sample.

Factors in determining the sample size:
purpose of the study
population size
allowable significance error
resources
Ideally, n 30 so that normality is assured. Normality is a requirement for most inferential
statistics.
Independent and Dependent Samples

Independent Samples (also called Unrelated Samples) - samples from different group of participants
are assigned to different experimental conditions (also called unrelated samples)
Example: 2 distinct sample groups, treatment and control groups
Dependent Samples (also called Paired, Matched or Related Samples) - samples from the same
group of participants are exposed to different experimental conditions.
Example: Pre-test and post-test scores for the same group of students
Statistical Hypothesis

is an inference about a population variable/ parameter that needs to be verified through data
collection, experimentation and testing.
an inference about a variable between two or more population
Examples: African people are not prone to cancer
Hybrid rice is better than native varieties in terms of yield
DOST employees are highly motivated workers

Hypothesis Testing
Process of verifying the validity of statistical hypothesis through the use of test statistics, i.e. ttest, chi-square, F-test, etc.
Computed value is compared to tabular value a given confidence level (1%, 3% or 5%), also
called level of significance.
If the computed value is greater than the tabular value, there is a significant difference in the
variables tested.

A computed probability value or p-value (also called significance value) of Statistical Programs
like SPSS can also be used as basis.
If p-value < 0.05 = significant
p-value < 0.01 = highly significant
p-value 0.05 = not significant
Page 5 of 11

STEPS IN DATA ANALYSIS

Data Entry

MS Excel may be used and data can be exported to another

software
First row should be variable names.
Input data in such a way that each row represents data from
one experimental unit while each column represents a
variable.
Example: if the entity is a person, each subject are assigned 1
row with all parameters obtained are allocated to the columns,
i.e., name, age, weight, etc.
Assign numerical codes for every categorical data, i.e. Sex
(Male = 1, Female = 2)

Visualization

Graphs and tables of summary statistics are excellent tools for visualizing data.
Identify any unexpected observations or patterns - which may due to errors in data collection or
data entry
Presence of data outliers can greatly affect the results of statistical tests

Page 6 of 11

Statistical Analysis Softwares

MS Excel can be used in basic statistical analysis

Statistical softwares, like SPSS, are much easier to use in data entry and analysis,
and it also contains advanced statistical tools.

CHOOSING AN APPROPRIATE TEST ANALYSIS

Test Statistics Given a Purpose/Objective
Purpose/Objective

Questions Answered

1. Descriptive

How many, proportion to overall

sample/population, average, range,
degree of dispersion of data, etc.

2. Correlational

Presence of, how strong and

direction of relationship (or
association) between uncontrolled
variables/groups

3. Experimental or
Inferential

3.1 Tests for

Differences of
Variables or
Groups
3.2 Comparison of
Group Means
(Post Hoc)

Existence of significant differences

between variables or groups;
Measures any change in one
variables when one or more
variables are manipulated.

Which specific variable or group has

significant difference over the others

Page 7 of 11

Common Analysis

Frequency, Percent, Cumulative

Count & Percent, Mean, Median,
Mode, Variance, Standard
Deviation, Range, CrossTabulations between
variables/groups, bar/pie charts

Pearson Correlation, Regression,

Contingency Coefficient, Cramers
V, etc.

ANOVA, Chi-Square Tests, MannWhitney U Test, McNemar Test,

Friedman Test, etc.
Tukey HSD, LSD, DMRT, etc.

Descriptive Analysis
Description Method

Nominal

Level of Measurements
Ordinal

Interval/Ratio

1. Frequencies

Count, Percent

2. Central Tendency

Mode

Mode, Median, Mean*

Mode, Median and Mean

4. Graphical Analysis

Pie, Bar

Pie, Bar

Scatterplot, Histogram,
Boxplots

3. Dispersion

Count, Percent,
Cumulative Count,
Cumulative Percent
Range

Examples:

Page 8 of 11

Count, Percent
Cumulative Count,
Cumulative Percent

Range, Percentile,
Variance and Standard
Deviation

Correlational Analysis for Two (2) Variables

Level of
Measurements

Variable Y

Nominal

Ordinal

For 2x2 Groups:

Pearsons Chi-Square Test

Variable X

Nominal

Ordinal

Interval/Ratio

(X ) or Fishers Exact Test

or McNemar Test
More than 2 Levels/Group:
Contingency Coefficient
(C), Cramers V (V) or Phi

Rank Biserial rrb

Point-Biserial
At least 3 levels of Nominal
Variable:
Eta Correlation

Rank Biserial rrb

Interval/Ratio
Point-Biserial
At least 3 levels of
Nominal Variable:
Eta Correlation

For 2x2 Groups:

Pearsons Chi-Square Test
2

(X ) or Fishers Exact Test

More than 2 Levels/Group:
Spearman Rho (r)
Kendalls Tau () (for small
sample)
Biserial rb

Biserial rb

Pearsons r (r)
Simple and Multiple
Linear Regression

Example 1:
Is there a significant relationship between Accountancy pre-board examination scores and actual
Accountancy board ratings of ESSU students?
Variable 1:
Level of Measurement:
Variable 2:
Level of Measurement:
Measure of Correlation:
Example 2:
Is shopping frequency (1st time, <1/mo., 1/mo., 1/wk, >1/wk) related to overall customer satisfaction
levels (strongly negative, negative, neutral, positive or strongly positive) in a mall?
Variable 1:
Level of Measurement:
Variable 2:
Level of Measurement:
Measure of Correlation:
Page 9 of 11

Experimental or Inferential Analysis

Test for Differences
Nominal

Level of Measurement: Dependent Variable

Ordinal

Interval/Ratio Scale

One Group:
One Sample Chi-square Goodness
of Fit Test (X2)

One Group:
Kolmogorov-Smirnov One Sample
Test (D)

One Group:
One Sample t-Test (t)

More Than Two-Groups:

Independent Samples: Chi-square
Test for Independence (X2)
Dependent/Paired Samples:
Cochrans Q Test (Q)

More Than Two-Groups:

Independent Samples: KruskallWallis One-Way Test or ANOVA
F-Test (F)
Dependent/Paired Samples:
Friedmans Test or Two-Way
ANOVA F-Test (F)

More Than Two-Groups:

Independent Samples: One-Way
CRD or Factorial ANOVA (F)
Dependent/Paired Samples: RCB or
Repeated Measures ANOVA (F)
Mixed Samples: ANOVA for
Differences or ANCOVA (F)

Two-Groups:
Independent Samples: Chi-square
Test for Independence (X2)
Dependent/Paired Samples:
McNemar Test (X2)

Two-Groups:
Independent Samples: MannWhitney U Test (U)
Dependent/Paired Samples:
Wilcoxon Signed-Ranked Test (T)

Two-Groups:
Independent Samples: Independent
Sample t-Test (t)
Dependent/Paired Samples: Paired
Sample t-Test (t)
Mixed Samples: ANCOVA (F)

Example 1:
Is there a significant difference between NAT results of pupils from public schools and private schools
in Borongan City?
Dependent Variable:
Level of Measurement:
Independent Variable:
No. of Groups Compared:
Type of Samples:
Appropriate Analysis:
Example 2:
Is there a significant difference between job satisfaction levels (strongly negative, negative, neutral,
positive or strongly positive) of workers from 4 Jollibee branches in Tacloban City?
Dependent Variable:
Level of Measurement:
Independent Variable:
No. of Groups Compared:
Type of Samples:
Appropriate Analysis:
Example 3:
Is there a significant difference between cholesterol levels of 20 patients before and after 6 months of
special diet?
Dependent Variable:
Level of Measurement:
Independent Variable:
No. of Groups Compared:
Type of Samples:
Appropriate Analysis:
Page 10 of 11

Comparison of Means (Post Hoc Tests for > 2 Groups)

Interval/Ratio Scale only

Homogeneity of Variance is Assumed:

Equal Sample Sizes: Tukey, LSD or Duncan
Slightly Different Sample Sizes: Gabriels Procedure
Sample Sizes are Very Different: Hochbergs GT2
Homogeneity of Variance is Not Assumed:
GamesHowell Procedure

Note: Levenes Test is used to determine homogeneity of variance. Significant result of the test
(p<0.05) means that the variances are not homogeneous

Example:

Is there a significant difference between the volume contents of 80 bottled beers (10 from each
machine) taken randomly from 8 bottling machines?
Dependent Variable:
Level of Measurement:
Independent Variable:
No. of Groups Compared:
Type of Samples:
Test for Differences:
Comparison of Means (assuming there is significant difference and there is homogeneity of
variance):

Page 11 of 11