CHAPTER 9:

“THE USE OF STATISTICS IN OUTCOMES ASSESSMENT”

I. General Concepts on Statistics and Learning Inquiry

Learning is more adequately achieved through research-based inquiry. Research-based inquiry

makes our academic pursuit scientific and therefore reliable, while guarding us against errors.
Through the ages, many things that people knew were proven untrue. Until the middle ages,
people thought the world was flat and the earth was the center of the universe, but the sciences
notably, navigation and astronomy, proved them false. As we look for truth in reality, science has
laid down criteria as solid bases for knowing based on empirical perception. So as human inquiry
became more conscious and rigorous, approaches and techniques were devised supported by
quantitative data and statistics.

In education, the use of statistics can cover much of the whole process of teaching and learning.
While this is not a book on statistics, it is relevant to make broad mention of the use of statistics,
particularly in an inquiry-directed or outcomes-based learning.

Definition of statistics

Statistics is a scientific discipline. It is a branch of Mathematics "that deals with the collection,
organization, presentation, computation and interpretation of data which are the, outcomes of
learning" (Santos et al, 2000). The statistician is an expert in the successful application of
statistical analysis or explanation of data. There are two general types of statistical analysis:

A. Descriptive statistics uses methods to summarize a collection of data by describing what was
observed using numbers or graphs. Numerical descriptions include mean and standard deviation
for data types (such as, weights and heights) or frequency and percentage for describing
categories (such as race and religion). This is most useful in communicating the results of an
inquiry, particularly experiments or surveys.
B. Inferential statistics, also called predictive statistics, uses methods to draw patterns in the
collected data, and then makes conclusions, predictions or forecasts about a group (also referred
to as population) or about a process being studied. The conclusions may take the form of (a)
answering yes or no questions (hypothesis testing); (b) estimating numerical characteristics of
data (estimation); (c) describing relations among data (correlation); (d) modeling relationships
among the data (for example, hypothesis analysis); or (e) predicting, forecasting or estimating
what has not yet been observed but which can be related to the population under study; (f)
extrapolation and interpolation of time series data or spatial data.

The meaning of statistics may be further exemplified through a number of relevant

characteristics.

1. Statistical inference is inductive moving from samples to the dimensions of a larger group of
a total population ("From individual interviews, 9 of 10 students in the class are from Metro-
Manila, therefore 90% or majority of students in the class are Metro-Manilans"). It is related, but
different from the Probability Theory which is deductive and moves from a total population to
deduce probabilities about samples (This is as if to say "I guess 9 out of ten students in the class
are Metro- Manilans; let's see how I can be proven right by asking the students their place of
residence").

2. Correlation does not imply causation. To exemplify, two characteristics or properties

(variables) of a population may be different, but are related. In studying household income and
mortality or death, for example, data may show that people of lower income die earlier than
those of higher income. The two variables (lower and higher income people) are said to be
correlated, although the cause of death may be different, such as by a non-considered factor, the
natural environment (called a confounding variable). If a non-considered factor exists, then a
causal relationship between the two considered variables may not be immediately established ("It
cannot be said that low household income causes household deaths").

3. Representative sampling is necessary to ensure that inferences Red conclusions can be

extended from the sample to the overall population. Statistics offers methods to correct random
sample trending within the procedures of data collection. Methods of experimental design can
also be used to lessen issues on correlation, while strengthening or making "robust" conclusions
on the study of a population. Note that some national political surveys are based on only about
2000 survey respondents.

4. Misuse of statistics can cause either errors in description or interpretation of data. Errors may
be so subtle that only experts can note the errors, serious enough as to cause grave errors in
conclusions affecting policies in governance, procedures in medical practice, or structures in
engineering projects. Hinting misuse of statistics, most politicians can therefore argue that
opinion survey results are inaccurate or wrong in case results are unfavorable to them. On the
other hand, survey experts say the margin of error in their surveys is too slim to be inaccurate or
wrong.

5. Results of correctly applied statistical techniques are difficult to interpret, and may need the
expertise of those who have intuitive sense to draw significant conclusions. However, it may
suffice to have statistical literacy or basic statistical skills which people can use in dealing with
information in their daily lives. Without being a statistician, the ordinary learner in school, for
example, can provide numerical interpretation of newspaper data results, such as on high rate in
H-fever cases. The ordinary student can mention reported numbers of those afflicted, and
compare these to numbers afflicted in previous months or year to draw conclusions on whether
there is a rise or fall in incidences of H-fever during the month or year.

Statistical methods

Statistics is useful in the teaching-learning process, along several research – based inquiries:

1. Experimental Studies

These inquiries investigate causes, in addition to drawing conclusions on the effect of changes in
elements (called variables) being studied. An actual experiment study, for example, can focus on
a class (called the experimental group) to which is applied a new method of learning, such as
technology integration in instruction. Through a set period, the learning achievement of the class
is accounted for and then compared with the achievement of another comparable class (called the
control group) which did not benefit from the new method of learning. Measurement and
comparison are made and the difference (variance) in learning achievement can be determined.
Results by way of a large numerical value of achievement can show that the new method of
learning is effective and can be formally introduced for regular classroom learning comparison
are made and the difference (variance) in learning achievement can be determined. Results by
way of a large numerical value of achievement can show that the new method of learning is
effective and can be formally introduced for regular classroom learning.

2. Inferential Studies

This does not involve any experimental group and control group. Instead, data are gathered and
the correlations between intervention (predictors) and the results derived from a single group are
investigated. For example, the correlation between smoking and lung cancer may be explored. A
survey may be used to collect observations followed by statistical analysis of collected data.

Generally, statistical inquiries observe five basic steps:

a) planning the research-based inquiry around size, hypothesis, variability, subjects, etc;

b) designing the experiment by blocking to reduce error; random assignment for unbiased
estimates, and mapping the procedures (experimental protocol);

c) implementation and analyzing data; d) further examination of data for secondary analysis; and

e) documentation and presentation of results of the study.

II. Conceptualization, Operationalization and Measurement

The use of statistics can help in the whole process of outcomes- based teaching and learning. If
the instructor, for example, wishes to teach concepts about abstract ideas such as patriotic love,
religious fanaticism, and political patronage, it is good to know that statistics can help clarify
such concepts. Ideas that the instructor wishes to communicate are actually idea-images, not the
physical or social realities themselves which are the subject of learning. This is the (or
phenomena) but concepts about external realities. This entails the process called
conceptualization.

Concepts as constructs
There are three classes of phenomena which can be statistically measured:

(a) direct observables, like color of a dress or heat of fire

(b) (b) Indirect observables which require indirect observations, such as the idea of a "female"
indirectly observed through signs of a person's gender
(c) (c) constructs or creations we form in the mind resulting from observations. An example of
these mind-constructs is intelligence quotient (IQ) which cannot be observed directly and
indirectly, but which educators have created. After assuming that IQ and other constructs
such as patriotic love or religious fanaticism have real meaning, we can begin to try to
understand their meanings, forming measurements of these constructs in order to better
understand, relate and make predictions about them.

Indicators and dimensions

An indicator is a sign of the presence of a concept (variable) under study. For example, visiting a
home for the aged can be an indicator of human compassion. Then we can assign numbers, as
indicators of levels or intensity of compassion to individuals, possibly judged by way of the
number of visits they made: 5 for Claire, 4 for Donna, 3 for Clarissa (girls); and 4 for Rafael, 3
for Justine and 2 for Jayron. We may calculate and conclude that the girls displayed an average
(mean) of 6.00, while the boys got an average (mean) of only 4.5. On the other basis of
quantitative analysis of group difference, we may conclude that girls are more compassionate
than boys. Dimension is a specific aspect of a concept combined into groups or sub-groups, such
as compassion toward neighbors/fellow nationals/foreigners/animals/plants. Various numerical
indicators can be assigned for each dimension. Then we can determine which group (neighbors,
animals, plants, etc.) the population-respondents have more compassion for.

Creating conceptual order

Concepts can become more clarified by a process of clarification called specification. This can
also be done by categorization or the ordering or ranking of data. For example, we may
categorize answers such as income in society by way of higher/middle/lower income levels or
P100,000/50,000/10,000 per month. The progression from our vague sense of what a term means
to specific measurements can be diagrammed as follows:
Conceptualization (process to specify what we mean)

Nominal definition (assigned to a term, not the real entity)

Operational definition (specifies how a concept is measured)

Real definition (better clarified status of a real thing)

Levels of Measurement

Variables are a logical set of attributes, e.g. gender. On the other hand, an attribute is a quality or
characteristic of something, e.g.. male, female. Attributes may represent any of the 4 levels of
measurement:

Nominal measure is a level of measurement describing a variable that has attributes which are
different, e.g. gender, birthplace, college major, etc. Male and female, for example, are distinct
and have no additional structures, thus they exhaust the possibilities of gender. This kind of
measurement simply counts and draws conclusions from the number of e.g. males, females,
birthplace, major subjects, etc.

Ordinal measure-a level of measurement describing a variable with attributes that can be in a
rank-order along some dimension. For example, socioeconomic status of communities can be
ranked along their attributes, such as their high, medium and low economic status.

Interval measure- a level of measurement describing a variable whose attributes are rank-
ordered and have equal distances between adjacent attributes, e.g. Fahrenheit temperature scale
in which the distance between 15 and 16 is the same as 87-88.

Ratio measure-a level of measurement describing a variable with attributes that have all the
qualities of nominal, ordinal and and based on a "true zero" point, e.g. age, length of residence in
a interval times in attending worship number of times divorced, and number of animal pets, etc.
Implications of the levels of measurement are relevant to the analysis of data. For example,
certain quantitative analysis techniques may require variables that meet certain minimum levels
of measurement, such as ordinal. And so, you should plan analytical techniques accordingly. It
may be noted, however, that the use of measures does not guarantee reliability. For example,
assessment tests established and used for a period of time may need fundamental revision due to
outdated topics or new trends in educational assessments.

III. Indexes, Scales, Typologies

Some learning inputs may not be easy to quantify, and may need added statistical techniques to
be understood. For example, the concept of age may require composite measures, equivalent to
asking several questions as an object of inquiry. The use of an index or scale may help to explain
and elicit understanding of concepts with a range of conceptual variations.

In an Index, scores for individual attributes are constructed. On the other hand in a scale, scores
are assigned to patterns of ideas. For example, civics can be signified through various actions
showing participation in political governance, such as: joining a tree planting activity, running
for a barangay position, supporting a socio-civic organization, etc.

On the other hand, a Scale is constructed by assigning scores to patterns of responses according
to higher and lower degrees of civic participation: 10 for running for public office, 7 joining pan-
government organization, 5 voting on election day, 0 for no participation at all.

Typology is the classification of observations in terms of attributes on two or more variables, an

example is the classification of broadcast networks as: commercial-urban, commercial-rural,
non-commercial- national, etc.

There are quantification techniques which the research-based learner can use to subject data or
concepts for better understanding, analysis or statistical interpretation. The simplest of
quantitative ing analysis is the univariate analysis, which involves one variable; the bivariate
analysis involves two variables, and: the multivariate analysis examines several variables
simultaneously such as age, education, and racial identity.

Quantification of data
Quantitative analysis can be handled by computer programs, as long as these computers are able
to read the data collected during research-inquiries. In case of a survey, some data are by
themselves numerical, such as age and income. Other data are easily quantifiable since numerical
representations can be assigned variables, such as "1" for male and "2" for female. Some data
may, however, be problematic, as these may require a coding process. This is especially true in
open-ended questions in which responses are non-numerical code to each reported occupation,
e.g. "1" manager, "2" craftsman, "3" artist, "4" non-skilled, etc.

Univariate Analysis

This is an analysis of a single variable for purpose of description. For example, gender can be
measured looking at how many of the respondents are men and women. The most basic format in
presenting univariate data is to show individual cases, listing the attribute for each case under
study.

Bivariate analysis

A bivariate relationship refers to two variables. For example, in measuring the respondents'
support for a student council organization for the Reproductive Health Bill, there are to
indicators of levels of support asked along two questions: (a) Do you feel the students support
the RH Bill is High, Moderate, Low, No Support at all? (b) Should the Student Council issue its
decision to support the RH Bill? The two items reflect degrees of support for the same thing with
(a) corresponding with (b). It may be expected that in case the support is overwhelming, the
support of the Student Council can be justified. Thus if opinions of support are met or not met,
there is a bivariate relationship between the two items.

Tests of statistical significance

Given the results of findings, there are inferential statistics which can assist in pursuading the
audience/readers as to the significance, stength, deserved interest in a completed research. These
are called parametric tests of significance. Short of stating the research is generally significant,
the tests refer to the likelihood that relationships between variables is significant enough to be
meaningful.

Other tests of statistical significance:

Chi Square. This is useful in social science and is based on the null hypothesis: the assumption
that there is no relationship between two variable, for example that there is no marked difference
between men and women favoring equality of the sexes.

An illustration of Chi Square: Issue on support for government program sterilizing men through
castration and women through tubal ligation in a rural setting.

An Illustration of Chi Square

Expected frequencies Men Women Total

Attended alumni affair 28 42 70
Did not attend affair 12 18 30
Total 40 60 100

Observed frequencies
Attended 20 50 70
Did not attend 20 10 30
Total 40 60 100
Observed minus Expected
Divided by Expected
Attended 2.29 1.52 x = 12.70
Did not attend 5.33 3.56 p = <.001

Path analysis
This assumes that the values of one variable ar caused by the value of another variable. This
makes the distinction between independent and dependent to be important.
Multivariate techniques. Other than the rudimentary statistical data manipulation, such as
through the use of tables and percentages, there are more elaborate models of statistical analysis.
For our purpose of orienting students to the use of statistics for the teaching-learning process, it
will suffice to provide some information on these more complex mode of data analysis.
There are other Multivariate Techniques, and it will serve our purpose to identify them. They are
as follows:
 a) Time-Series Analysis. This represents changes in one or more variables over time. The crime
rate in society, for example, can be expressed along time fluctuations allowing prediction of
crime w in coming years.
b) Factor Analysis. This is used to discover patterns among the variations in values of several
factors Using a complex algebraic method, the problem of delinquency can be studied using
questionnaire which asks respondents, e.g. young people to report if they have committed
delinquent acts according to types of delinquent acts. Responses are then subjected to factor
analysis.
c) Analysis of variance (ANOVA). Cases under study art combined into groups representing an
independent variable, and the extent to which the groups differ from one another is analyzed in
relation to some dependent variable. The One-way Analysis of Variance can study, for example,
income levels of households in the Dasmarinas Village and Corinthian Gardens Subdivision.
Calculations may be performed by computer programs, simplest of which is known as "t-test"
which determines the difference between two means (averages.) On the other hand, the Two-way
Analysis of Variance allows the simultaneous study of more than two variables. For example, in
determining the difference in household incomes of two villages, the function of education may
be included. Possible result may be that more educated households get higher income.
Other multivariate techniques are the Discriminant Analysis, Log Linear Models and the
Geographic Information systems which require more sophisticated statistical procedures which
can be learned in Statistics or formal Research classes.