Statistics 

is the study of the collection, organization, analysis, interpretation and presentation of data to
yield meaningful information.  It deals with all aspects of data, including the planning of data collection in
terms of the design of surveys and experiments.

Descriptive and Inferential Statistics

 Descriptive Statistics

Descriptive statistics comprises the kind of analyses we use when we want to describe the population
we are studying, and when we have a population that is small enough to permit our including every
case. It deals with the methods of summarizing and presenting a mass of data so as to yield meaningful
information.

For example, we might want to describe a physics class and compare it to a class of English literature.
We might want to compare the gender composition of the classes, the math attitudes of the two
classes, the familial demands and supports of the two classes. Descriptive statistics would allow us to
do this.

 The classes are small enough that we can include the whole class in our studies.
 We would have to agree on how to measure gender, math attitude, and familial demands.
 We could then interview the students in the classes, or survey them, or ask them to tell their
stories about these issues and code them by content analysis.
 Then we COULD DESCRIBE how these issues affect the members of the classes we studied, and
how these variables are related in those classes.
 We COULD NOT CONCLUDE that our results could be generalized to all physics and english
literature classes because we have no idea whether the classes in our study were
REPRESENTATIVE OF all physics and English literature classes.

 Inferential Statistics

Deals with making generalizations about a body of data where only a part is examined. This comprises
those methods with the analysis of a subset of data leading to predictions or inferences about the
entire set of data.

The important keys to the difference between descriptive and inferential statistics are the capitalized
words in the description: COULD DESCRIBE, COULD NOT CONCLUDE, AND REPRESENTATIVE OF.

Descriptive statistics can describe the actual sample you study. But to extend your conclusions to a
broader population, like all such classes, all workers, all women, you must be use inferential statistics,
which means you have to be sure the sample you study is representative of the group you want to
generalize to.

This means you can't do a study at the local mall and claim that what you find is valid for all shoppers
and malls.

You can't do a study on college sophomores and claim that what you find is valid for the general
population.

You can't give a women's movement that includes a majority of a single ethnic group and claim that
what you find is valid for women of all ethnic groups.

As you can see, descriptive statistics are useful and serviceable if you don't need to extend your results
to whole segments of the population. But the social sciences tend to esteem studies that give us more
or less "universal" truths, or at least truths that apply to large segments of the population, like all
teenagers, all parents, all women, all perpetrators, all victims, or a fairly large segment of such groups.

Leaving aside the philosophical and methodological soundness of such a search for some kind of
general conclusion, different statistical approaches must be used if you aspire to generalize. And the
primary difference is that of SAMPLING. You must choose a sample that is REPRESENTATIVE OF THE
GROUP TO WHICH YOU PLAN TO GENERALIZE.
What are quantitative and qualitative data?

Quantitative data are measures of values or counts and are expressed as numbers.

Quantitative data are data about numeric variables (e.g. how many; how much; or how often). 

Qualitative data are measures of 'types' and may be represented by a name, symbol, or a number code.

Qualitative data are data about categorical variables (e.g. what type).

Quantitative = Quantity Qualitative = Quality

Data collected about a numeric variable will always be quantitative and data collected about a categorical variable
will always be qualitative. Therefore, you can identify the type of data, prior to collection, based on whether the
variable is numeric or categorical.

Why are quantitative and qualitative data important?

Quantitative and qualitative data provide different outcomes, and are often used together to get a full picture of
a population. For example, if data are collected on annual income (quantitative), occupation data (qualitative)
could also be gathered to get more detail on the average annual income for each type of occupation. 

Quantitative and qualitative data can be gathered from the same data unit depending on whether the variable of
interest is numerical or categorical. For example: 

Data unit Numeric variable = Quantitative data Categorical variable = Qualitative data
A person "How many children do 4 children "In which country were Australia
you have?" your children born?"
"How much do you earn?" $60,000 p.a. "What is your Photographer
occupation?"
"How many hours do you 38 hours per week "Do you work full-time or Full-time
work?" part-time?"
A house "How many square metres 200 square metres "In which city or town is Brisbane
is the house?" the house located?"
A business "How many workers are 264 employees "What is the industry of Retail
currently employed?" the business?"
A farm "How many milk cows are 36 cows "What is the Dairy
located on the farm? main activity of the farm?"

How can you use quantitative and qualitative data?

It is important to identify whether the data are quantitative or qualitative as this affects the statistics that can be
produced.

Frequency counts:

The number of times an observation occurs (frequency) for a data item (variable) can be shown for both
quantitative and qualitative data.

The graphs below arrange the quantitative and qualitative data to show the frequency distribution of the data.
Quantittive Data

Qualitative Data

Classification of Quantitative Variable

Discrete data
When the values in the batch are whole numbers (counts), the data set is called discrete. Examples of
discrete measurements are:

 the number of admissions in a hospital's accident and emergency unit each day over a period of
two months,

 the number of people in each household in a survey of 10,000 households,

 the number of broken eggs in each of 200 boxes of 1 dozen eggs arriving at a supermarket,

 the number of issues of a particular magazine purchased over the previous 12 months by each
respondent to a telephone survey.
 Continuous data
When the data are not constrained to be whole numbers, the data set is called continuous. Results from
infinitely many possible values that can be associated with points on a continuous scale. Examples are:

 the maximum temperatures each day in January in your local city,

 the daily calorie intake of each child in a health survey,

 the daily exchange rate between the USdollar and Peso over the past month.

 Grade point average

 

What is Level of Measurement?

In science, applying measurements appropriately is considered as the key skill for collating life science data.
These measurements come in wide forms and require varied approaches for optimization. Some variables are
continuous(or metric), which fall alone on an uninterrupted scale(e.g. length, mass, temperature) while count
data can only be whole numbers and are categorical (also termed discrete, discontinuous or non-numeric).

Four levels of measurement are recognized as follows:

1. Nominal
2. Ordinal
3. Interval
4. Ratio Scales

Level of Measurement in Research

There are four levels of measurements used in research. Often the quantitative research does not feature the
sharp boundaries as found in quantitative research in relation to devolping measures, collecting data with those
measures, and evaluating the measures. Level of measurement refers to the way that a variable is measured.

Nominal Scale

A nominal scale classifies observations into exclusive categories that have no relative rank. The mode of the
nominal attribute, which gives its central tendency mean and median, cannot be defined here. Nominal scale uses
labels. The variables on the nominal scale are known as categorical variables. It has a unique way of naming the
attributes.

Example: 
Species, gender, color and habitat type.

Ordinal Scale

An ordinal scale classifies observations into ranked exclusive categories where attributes are ordered according to
the rank. The distance between the attributes of relave size does not play any role here. The rank order of the
entities is represented by the number of objects or events. The mode or median describes the central tendency;
mean cannot be defined here.

Example: 
Abundance scales (e.g., the DAFOR scale (dominant, abundant, frequent, occasional, rare)) used to record the
abundance of plant species in a quadrat and development state (e.g, newborn, juvenile, yearling, adult).
Interval Scale

An interval scale defines observations on a continuous scale that has no absolute zero. All types of quantitative
attributes can be measured in interval scale. The variables which can be measured using this scale are known as
scaled variables or interval variables. The ratio between the numbers in the scale are not same. Its mode or
median or arithmetic median represents the central tendency of the interval scale.

Example: 
The Celsius temperature scale and the calendar date. (Note that the absence of an absolute zero means, it is not
true to say that 15oC is three times warmer than 5 oC, nor that birds arriving on the 8 thmay are twice as late as those
arriving on 4th may.)

Ratio Scale

A ratio scale defines observation on a continuous scale that has an absolute zero. We can find a meaningful ratio
for the ratio scale. In the fields like engineering and science, we widely use ratio scale. All mathematical operations
can be done in the ratio scale. It also can represent central tendency with all types of mean and median. The
measurement that requires ratio can also be represented in ratio scale.

Example: 
Length, weight and most physical measurements. Thus, in contrast to measurements made in the interval scale, it
is true to say that a plant leaf of length 45mm is half the length of 90mm length.

Likert Scale Level of Measurement

Likert Scale is one kind of rating scale that has been developed to measure attitudes directly. Likert Scale is a five
(or seven) point scale which is used to allow the individual to express how much they agree or disagree with a
particular statement. The response categories in Likert scales have a rank order, but the intervals between values
cannot be presumed equal. 

In the four level of measurements, the ordinal, interval and ratio levels of measurements are more exact and clear
when compared to the nominal data of measurement. In the nominal measurement of data, we are just classifying
the data. The data at all the other three levels has the properties to compute at least the basic statistical measures.
So, we can say that the three levels of data measurement, ordinal, interval and ratio play a vital role in Statistics.

To summarize, we can say :

 Nominal measurement - Based on attributes.

 Ordinal measurement - Based on ranks.
 Interval measurement - Based on order with no meaningful 0.
 Ratio measurement - Based on order with a meaningful 0.

Nominal Level of Measurement

Nominal measurement is the easiest type of the measurement of data. In this case, we just have to identify the
category. The word nominal is derived from the Latin word “Nomen” which means, “Name”. Nominal data is also
known as qualitative data, categorical data or attribute. It reflects the qualitative differences rather than the
quantitative differences.

While dividing data as nominal, we should check whether the data is mutually exclusive and exhaustive. Mutually
exclusive means, the occurrence of one prevents the occurrence of other and exhaustive means, exactly one of
them can occur. In simple words, if we measure a data nominally, each item can come under one and exactly one
head.
Examples of Nominal Level of Measurement

Examples of nominal level measurement are given below:

1. Eye color of the contestants in a contest is considered as nominal data or attribute as one person can have
exactly one and only one eye color (say brown). He cannot have more than one eye colors. So, the person
can be characterized by his eye color. We can check whether a person has Brown eye color or not.
2. Intelligence, beauty etc are also considered as nominal data or attribute as we can say whether a person
possess beauty or intelligence.

(i) We can use words and letters to represent nominal measurement of data. Suppose, there are two categories
namely men and women, we can represent them as M and W so that, all men come under the category M and all
women come under the category W. We can see that category name is the only thing matters here.

(ii) With nominal data, we can only do the mathematical operations like counting and simple statistical operations
such as mode.

Ordinal Level of Measurement

Ordinal measurement is based on ranking of data. This measurement orders the items using the ranks. There is no
constant distance or difference between the items, in order, in this case.

Ordinal Level of Measurement Examples

Some examples of ordinal level measurement are given below:

1. The student who scores the highest in a class is assigned the rank 1. The student, who scores second
highest is assigned the rank 2 and so on. The difference between the marks scored by the student with
first and second rank may be 10, where as, the difference between the marks scored by second and third
rank may be 12. So, we can see that, there is no common difference between the marks obtained by the
students. The students are ordered based on the marks they scored.
2. If we are asked a question, “How often you go to Super Market?”, with the four answers to choose from
“a. Frequently, b. Sometimes, c. Rarely, d. Never” Here, there is no constant difference between 1, 2, 3
and 4. We are just arranging the people based on the frequency of their visit to Super Market.

Like Nominal data, the ordinal data also lacks the properties which are required to compute the statistical
measures like average.

Interval Level of Measurement

Interval measurement of data is the measurement based on both ranking and regular intervals between scale
points. It not only orders the measurement, but also clearly defines the distance between the orders.
Interval Level of Measurement Examples

Some Examples of interval level of measurement are given below:

1. The difference between the intervals 20 - 30 is same as the difference between 50 - 60. In both cases, the
difference is 10 and also we know that 50 - 60 is higher than 20 - 30. So in this case, we know both the
order and distance between the intervals.
2. The room temperature of 5∘−50∘ is same as the room temperature of 15∘−25∘. Here also, we know
both the order and also distance between the intervals.

(i) Ratios are not meaningful in the case of interval data.

(ii) There will be no zero in the interval data.

Ratio Level of Measurement

Ratio measurement is very similar to the interval measurement. When we include a meaningful zero in the interval
measurement, it becomes ratio measurement. Ratio measurement of data is the strongest of all levels of
measurement of data. Zero is considered as the point of reference in this type of measurement of data. It is not
compulsory that zero is a value of the items in this case. In the ratio level of measurement, the distance between
the orders on the scale are equal, and the ranks are assigned to the items, according to their size or magnitudes.

Ratio Level of Measurement Examples

Some examples of ratio level measurement are given below:

1. A company has zero profit, it means, it has no excess amount than it spent. So, here we have a meaningful
zero, which denotes no profit. If we say the company has 100 dollar profit, it has 10 dollar more than zero
profit.
2. If we say a baby of one year is 1.5 times heavier than a newly born baby. Here baby's weight is clearly a
ratio data. We know that a baby cannot have zero weight. But, the ratio is meaningful.

1. In ratio measurement of data, zero is the basis of the measurement.

2. Ratio data has all the properties of the data of the other three levels of measurement.
 Summary of The Four Levels of Measurement: 
Appropriate Descriptive Statistics and Graphs

James Neill, 2009

Level of Properties Examples Descriptive Graphs

Measurement statistics

Nominal / Discrete Dichotomous Frequencies Bar

Categorical Arbitrary   Yes / No Percentage Pie
(no order)  Gender Mode
Types / Categories
 colour
 shape

Ordinal / Rank Ordered Ranking of favourites Frequencies Bar

categories Academic grades Mode Pie
Ranks Median Stem & leaf
Percentiles

Interval Equal distances Discrete Frequencies Bar

between values - Thoughts, (if discrete) (if discrete)
Discrete  behaviours, feelings, Mode Pie
(e.g., Likert scale) etc. on a Likert scale (if discrete) (if discrete)
Metric  Metric Median Stem & Leaf
(e.g., deg. F) - Deg. C or F Mean Boxplot
Interval scales >5 SD Histogram
can usually be Skewness (if metric)
treated as ratio Kurtosis

Ratio Continuous / Age Mean Histogram

Metric / Weight SD Boxplot
Meaningful 0 VO2 max Skewness Stem&Leaf
allows ratio Deg. Kelvin Kurtosis (may need to
statements round leafs)
(e.g., A is twice as
large as B)

Summary of Descriptive Statistics & Graphical Summaries for the Four Levels of Measurement

Statistic Nominal Ordinal Interval Ratio

Mode √ √ √ If meaningful
Median X √ √ √
Range, Min. Max X √ √ √
Mean X X If metric √
SD X X If metric √
 
Graph Nominal Ordinal Interval Ratio
Bar / Pie √ √ If discrete X
Stem & Leaf X √ √ √
Boxplot X √ √ √
Histogram X X If metric √