Q

Quantitative Methods quantify effect sizes, determine the strength of

associations, rank priorities, and weigh the
Juwel Rana1,2,3, Patricia Luna Gutierrez4 and strength of evidence of effectiveness.
John C. Oldroyd5
1
Department of Public Health, School of Health
and Life Sciences, North South University, Introduction
Dhaka, Bangladesh
2
Department of Biostatistics and Epidemiology, This entry aims to introduce the most common
School of Health and Health Sciences, University ways to use numbers and statistics to describe
of Massachusetts Amherst, MA, USA variables, establish relationships among variables,
3
Department of Research and Innovation, South and build numerical understanding of a topic. In
Asia Institute for Social Transformation (SAIST), general, the quantitative research process uses a
Dhaka, Bangladesh deductive approach (Neuman 2014; Leavy 2017),
4
Independent Researcher, Masatepe, Nicaragua extrapolating from a particular case to the general
5
School of Behavioral and Health Sciences, situation (Babones 2016).
Australian Catholic University, Fitzroy, VIC, In practical ways, quantitative methods are an
Australia approach to studying a research topic. In research,
the investigator needs to choose specific tech-
niques to collect, analyze, and interpret data
Synonyms according to the problem’s nature or the research
question under study. For example, a researcher
Quantitative analysis; Quantitative research may want to know the top ten leading causes of
methods; Study design diseases affecting a country. The most appropriate
way of addressing this is by using data from a
national survey, one of the most common instru-
Definition ments for primary data collection and secondary
analysis in quantitative research. In this case, a
Quantitative method is the collection and analysis national health survey provides quantitative data
of numerical data to answer scientific research on the scope and distribution of illnesses at a
questions. Quantitative method is used to summa- national level by studying a representative sample
rize, average, find patterns, make predictions, and of the whole population. However, it will be hard
test causal associations as well as generalizing to understand people’s challenges with visual
results to wider populations. It allows us to impairment when trying to access public services
© Springer Nature Switzerland AG 2021
A. Farazmand (ed.), Global Encyclopedia of Public Administration, Public Policy, and Governance,
https://doi.org/10.1007/978-3-319-31816-5_460-1
2 Quantitative Methods

with a quantitative survey. In this case, applying a representative sample size. This aspect is also
qualitative method for data collection and analysis essential if the study requires specific charac-
can provide a better understanding of their expe- teristics to be present in the sample, such as
riences and challenges. having the same number of male and female
Some examples of research questions that can participants.
be answered using quantitative methods are: Data collection: There are many ways in which
survey data can be collected: Internet, tele-
• What proportion of people were born outside phone, using regular mail, or face to face
of a country? (Treiman 2009). The method chosen will
• How much rainfall was there in a region in depend on the time and resources available.
2020? Response rate: Survey response rates are often
• Does a strict lockdown prevent new cases of low. A sampling protocol needs to establish
COVID-19? ways to direct the survey to the right respon-
• Is drug A more effective than drug B in pre- dents (the target group) to gain a greater
venting liver cancer from hepatitis C? response and generate more accurate data.
Sampling technique: The sampling protocol
In order to answer these questions, the should also specify a sampling technique: sim-
researcher must choose an appropriate method of ple random (every individual has the same
enquiry. probability of being selected to take part in
the study), systematic (participants are selected
in defined criteria), cluster (instead of just
looking at individuals, the population is
Methods of Inquiry divided into groups (e.g., schools, local gov-
ernment areas) which represent a unit of study
Methods in quantitative research can broadly be
that latter can be randomly selected to be
classified as observational or experimental.
included in the study), and stratified
(participants are divided into groups with sim-
ilar characteristics) (Brenner 2020).
Observational Research Method: Survey design: There are two survey designs:
Quantitative Survey cross-sectional and longitudinal (Neuman
2014). The first one collects information at
The survey is a common technique in quantitative one point in time, and the latter collects infor-
observational research (Creswell 2014); it pro- mation at multiple points in time. There are
vides a numeric description of sample character- three types of longitudinal studies: cohort
istics or the whole population under study (follows the same population across time, usu-
(Balnaves and Caputi 2001). It is widely used in ally starting at birth), trend (like a cross-
a variety of fields to generate information. When sectional survey, but repeated multiple times
working with data from a representative sample, to different people), and panel (ask the same
the results can be generalized to the entire popu- people, the same questions over time).
lation. Surveys usually consist of standardized Survey instrument: Depending on the nature of
questions analyzed using descriptive and inferen- the topic and the resources available, the
tial statistical tools. When working with surveys, researcher can choose to use an existing instru-
it is necessary to take into consideration the ment, a modified instrument, or design a new
following: instrument. In all the cases, the instrument
should be tested for validity (instrument mea-
Aim: Have a clear purpose for the survey. sures what it is supposed to measure) and reli-
Characteristics of the population: Knowing the ability (instrument is consistent between
whole population’s size is critical to determine measurements) (Creswell 2014).
Quantitative Methods 3

Experimental Research Method cannot allocate participants randomly, usually

due to ethical reasons.
Experimental research is the oldest form of quan- Confounders: Confounder is the common cause
titative research (Creswell 2014). It aims to iden- of exposure and outcome. In other words, con-
tify relationships between variables by measuring founders are additional variables that influence
direct observations. Therefore, it requires obser- both independent and dependent variables.
vation and testing of the cause-effect relationship Confounders need to be taken into consider-
between two or more variables (dependent and ation. Otherwise, they may mask true effects.
independent). In a laboratory setting, the indepen- For example, in examining the association
dent variable is often called the “exposure.” It can between smoking and lung cancer, researchers
be a new drug or a vaccine. The dependent vari- need to be aware that alcohol consumption is a
able is an outcome that “depends” on the expo- confounder. This is because people who smoke
sure. It can be the infection rate, cell proliferation, are more likely to drink alcohol, and alcohol
or enzyme activity. Outside of a laboratory set- may also influence lung cancer. Without con-
ting, the exposure can be the habit of smoking sidering the confounding effect of alcohol, we
cigarettes or performing a specific job type. Expo- may be misled into concluding that alcohol
sure is a characteristic that is being manipulated or consumption causes lung cancer, not smoking.
assigned for the study. Here we say alcohol consumption “confounds”
The dependent variable is the outcome vari- the (true) association between smoking and
ables influenced by the independent/exposure var- lung cancer. Researchers need to account for
iable and confounders. The outcome variable is confounding variables to infer accurate
observed and measured but is not manipulated. It conclusions.
is usual in experimental design to have a second or Control group or placebo: A control group is a
third arm, where a different type of intervention or group with the same characteristics as the inter-
no intervention/placebo is being applied. How- vention groups at the start of the study but
ever, this is an oversimplification of the reality receives placebo/no intervention. This allows
when conducting experiments. When doing the researcher to contrast the outcome between
experimental research, it is necessary to take into those who do and those who do not receive the
consideration a wide variety of aspects: intervention/treatment.

Hypothesis: We should keep in mind that exper-

imental research’s main objective is hypothesis
Data Analysis Techniques
testing. However, to generate a well-sounded
hypothesis, it is important to understand previ-
Data analysis depends on the study design and
ous literature and theories. Based on that
objectives of the study. There are two broad
knowledge, a new hypothesis can be formu-
ways to approach data analysis: descriptive and
lated and tested.
inferential statistics.
Experimental design: There are two broad ways
of conducting experimental study: randomized
controlled trial and quasi-experimental study.
The first one is also known as the gold standard Descriptive Statistics
experimental design, where the participants are
randomly allocated to the experimental treat- Descriptive statistics summarize information and
ment or intervention or a control/placebo arm provide a numerical description of the data. There
where there is no intervention. Randomized are three ways in which data can be described:
controlled trial is viewed as strong causal frequency, measures of central tendency, and
study design. The second one is also known using measures of dispersion.
as causal comparative, where the researcher
4 Quantitative Methods

Frequency value of the observations is subtracted from

the highest value.
Frequency provides the number of observations IQR: Following the idea on the range, IQR is the
under a category. Usually, it is presented as a difference between the 75th quartile and the
percentage. For example, we may describe boys’ 25th quartile.
and girls’ frequency who play soccer in school. SD: SD is the most common way to measure
dispersion. It measures how far the data spread
from the mean. If the SD is small, the data is
close to the mean. On the contrary, if the SD is
Measures of Central Tendency
large, the data is spread away from the mean.
SD is widely used in hypothesis testing.
These types of measures have the characteristic of
Z-score: Z-score is a combined measure con-
summing information. There are three measures
structed using an SD and the mean, allowing
to describe central tendency: mode, median, and
comparisons among groups and distributions.
mean. When all of these measures of central ten-
dencies match, the data set is considered to follow
a normal distribution.
Inferential Statistics
Mode: Mode represents the most common obser-
vation. For example, ten people are weighted: Inferential statistics go beyond description by
four of them weight 75 kgs, and the rest of the aiming to estimate the strength and direction of
participants weight 72, 66, 73, 76, 100, and relationships between variables and make infer-
71 kgs. In this case, the mode is 75 kgs. ences about a larger group using data from a
Median: The value in the middle of a set of representative sample of that group (Babones
numbers is organized from lowest to highest. 2016). However, researchers need to be confident
Median is also defined as the middle point about these associations’ strength since there is
because 50% of the data is above this point always chance of error. One of the main pillars
and the other 50% is below. of inferential statistics is hypothesis testing.
Mean: Mean represents the average of the obser-
vations. To calculate the mean, all the observa-
tions are added up and then divided by the
Developing Hypothesis for Testing
number of observations. It is important to
keep in mind that the mean is affected by
Hypothesis testing can be confusing because the
extreme values.
researcher must assume no difference or relation-
ship between variables until proven to the con-
trary. This process involves creating two
Measures of Dispersion hypotheses: a null hypothesis and an alternative
hypothesis. The null hypothesis (Ho) always states
These measures provide information on the vari- that there is no difference between one group
ability within the data, which is not obtained using (A) and another (B). The notation is Ho: A ¼ B.
the measures for central tendency. There are four The alternative hypothesis (Ha) states the opposite
common ways to measure dispersion: range, inter- that there is a difference between one group
quartile range (IQR), standard deviation (SD), and (A) and another (B) (Ha: A 6¼ B). The idea is to
Z-score prove the null hypothesis wrong or rejecting the
null hypotheses. Some factors need to be consid-
Range: The range is the difference between the ered to make a decision, such as the significance
extreme values: the minimum and maximum level, p-values, and types of errors (Type I and
value. The calculation is simple; the lowest Type II).
Quantitative Methods 5

Significance Level (Alpha) Tests in Inferential Statistics

Alpha (α) refers to the probability of rejecting the There are several tests in inferential statistics.
null hypotheses when it is true. You say there is a Choosing the most appropriate test will depend
difference between groups when there is no dif- on the nature of the data and the study objective
ference. The usual cutoff point for level of signif- (Mertens et al. 2018). Some of the most common
icance is set at 0.05. It can be interpreted as the tests are mention below:
probability (5%) of inappropriately rejecting the
null hypothesis, which is known as a Type I error. • Chi-square: Chi-square test examines the
association between two categorical variables.
• Spearman correlation: Spearman correlation
tests the association of two ordinal variables.
Statistical Significance (P-value)
• Pearson correlation: Pearson correlation
measures the association between two contin-
Statistical significance is used to ensure that the
uous variables.
differences observed between two sample groups
• Paired t-test: Paired t-test examines the differ-
(experimental and control) are not due to chance
ence between two variables in the same indi-
alone; therefore, the difference observed is a real
viduals of a population (e.g., test scores before
difference between the two groups. It is called
and after an intervention).
P-value, and it can take any number between
• Independent t-test: Independent t-test exam-
zero and one. Putting it in words, the P-value is
ines the difference between two variables from
the likelihood that the observed difference
different groups. For example, when analyzing
between groups is a finding by chance. The
a characteristic to see if it is different between
smaller the P-value, the greater our confidence
boys and girls.
that this is not a finding by chance and correct to
• Analysis of variance (ANOVA): ANOVA is
reject the null hypothesis. By convention, if the
used to compare the difference of two or more
P-value is less than 0.05, the null hypothesis can
means from different groups.
be rejected. It can be interpreted as having 95%
• Simple regression: Simple regression esti-
probability that the results are not due to chance
mates how much dependent variable changes
alone.
when independent variable change (e.g., how
much blood pressure changes when body mass
index changes).
Errors: Type I and Type II • Multiple regression: Multiple regression esti-
mates the changes in dependent variable based
The cutoff points established for the Alpha (α) and on multiple explanatory variablesor indepen-
P-value are low; however, there is always a 5% dent and confounding variables.
chance of making an error. In hypothesis testing,
we can have two types of errors: Type I and Type
II (Neuman 2014). A Type I error is when a true
Interpretation
null hypothesis is rejected (also defined as a false
positive), which is a more problematic type of
To wrap up, data interpretation requires going
error. A Type II error is when the researcher
back to the beginning and looking at the initial
accepts a false null hypothesis. This type of error
research question, observing new insights from,
is influenced by sample size and variability. It is
and drawing conclusions. Usually, these conclu-
less problematic than Type I error.
sions will have substantial policy implications or
impacts on decision making.
6 Quantitative Methods

Conclusion References

Simply, quantitative research methods are system- Babones S (2016) Interpretive quantitative methods for the
social sciences. Sociology. https://doi.org/10.1177/
atic ways of gathering and analyzing numerical
0038038515583637
data with the general purpose of understanding Balnaves M, Caputi P (2001) Introduction to quantitative
reality. Quantitative methods are mostly used to research methods: an investigative approach. Sage,
generalize results to a wider population, make London
Brenner PS (2020) Understanding survey methodology:
predictions, infer causality, and prove hypotheses.
sociological theory and applications. Springer, Boston
Creswell JW (2014) Research design: qualitative, quanti-
tative, and mixed methods approaches. Sage, London
Cross-References Leavy P (2017) Research design. The Gilford Press,
New York
Mertens W, Pugliese A, Recker J (2018) Quantitative data
▶ Dependent Variables analysis, research methods: information, systems, and
▶ Empirical Research contexts: second edition. https://doi.org/10.1016/
▶ Experimental Design B978-0-08-102220-7.00018-2
Neuman LW (2014) Social research methods: qualitative
▶ Hypothesis Testing
and quantitative approaches. Pearson Education Lim-
▶ Measure of Central Tendency ited, Edinburgh
▶ Measure of Dispersion Treiman DJ (2009) Quantitative data analysis: doing social
▶ Research Design research to test ideas. Jossey-Bass, San Francisco
▶ Research Method
▶ Quasi-Experimental Design