Survey- Correlation

SUMMARY OF DESCRIPTIVE QUANTITATIVE (SURVEY) AND CORRELATIONAL RESEARCH

       

Quantitative research covers some designs of research, two of which, descriptive and correlational, will described in this summary. Along with the description of the two designs, the types and characteristics as well as the examples for clearer understanding are provided.

A.    Survey Research

Survey research is also named descriptive quantitative research. As the name suggests, this type of quantitative research aims at describing individuals, groups, institutions, methods, or materials, which from the description, the researchers can compare, contrast, classify, analyze, and interpret the entities and the events (Cohen, 2007). Best (1970) defines what is described in the survey research as conditions, beliefs, point of view, attitudes, process of how something is going on, effects felt by certain groups, and trends which are developing. To be specific, a researcher would be likely to find the answer to how the observed subject related to some preceding event that has influenced or affected a present condition or event.

1.      Types of Survey Research

To classify types of survey research, there are two divisions; a) according to focus and scope; and b) according to time dimension for data collection. Each of the divisions possesses its own categories or types. Therefore, to provide clearer description of each of the categories, the following paragraphs are present.

Before going further to the categories of type of survey according to the scope and focus, it is important to know that scope uses terms like ‘census’ and ‘sample survey’, while focus takes ‘tangible’ and ‘intangible’ terms. In the view of scope of the survey, ‘census’ means that there is no need to draw sample because all participants are the source of the data, while the term ‘sample survey’ implies that the population is so large that researchers need to draw samples. Besides, ‘tangible’ and ‘intangible’ are used to refer to the concreteness of the information being sought.

a.      A Census of Tangibles

When a research’s scope is a small population and it focuses on seeking concrete information, such survey is a census of tangibles.

b.      A Census of Intangibles

The scope of the research in this type of survey is a small population; there is no need for the researchers to draw samples. The focus, however, is abstract information which data can be collected through tests, questionnaires, or other instruments.

c.       A Sample Survey of Tangibles

The type of survey research is when a research takes a big population, from which the researchers have no access to the whole population; sampling is needed, while the information being sought is concrete or tangible.

d.      A Sample Survey of Intangibles

In this type of survey, the population is also big, thus sample is drawn, and the focus or information sought is intangible.

From the second division of type of survey is based on the time dimension for data collection, two categories emerge, namely longitudinal and cross-sectional survey research. The following provides further explanation of the two types. 

a.      Longitudinal

Its purpose is to describe changes of single sample over extended period of time. This type of survey takes a long time (therefore it is longitudinal) because what is studied is the long-term changes over any certain variables. Therefore, longitudinal studies are usually able to find the complexity of human behavior. This type of survey can be done ongoing or use repeated cross-sectional studies.

            There are three kinds of longitudinal study namely panel, trend, and cohort. Derived from the main definition of longitudinal study, which is a study conducted in a period of time to see changes, the three kinds to be explained mainly differ in the subject or sample used during the study.

                        Panel Study

In this kind of longitudinal survey, the same individuals are observed over time. If the technique for data collection is interview, the same individual is interviewed from time to time regarding the same problem in order to note changes. Since the sample is the same, any changes observed are not a result of sampling error. An example of study employing this design is that a study of learning style of junior high school students starting in their first year. In such study, a researcher will take a group of students with some common characteristics and observe them from the first year until the final year in junior high school.

Cohort Study

For cohort study, a particular population is followed with different samples that are taken randomly. Some references mention cohort as the same as panel, however. What differs is the sampling, in which the panel study uses the same sample over time and cohort uses different samples for every observation, but still from the certain population. Cohen (2007) mentions in his book that an example for cohort study is the survey of National Child Development Study. A different representative sample was interviewed each year, which was taken randomly, started from 1958.

Trend study

This trend study examines overall change over time because the focus of trend is on factors from which can be drawn pattern to be used as prediction of trend in the future. Knowing the focus is factors, ‘new samples are drawn at each stage of the data collection…’ (Cohen: 2007). New samples, however, should possess certain characteristics so that changes can be validated. In addition, some trend studies may find unpredicted factors that may cause the formulated pattern invalid. An example to this kind of longitudinal study is that the observation of university students’ educational progress.

b.   Cross-sectional

It is described as a type of survey which studies a certain variable at a certain time.  This type of survey studies the characteristics of a variable at one precise time, for example, the certain age of learners is studied to know learning characteristics so that the findings can be used as supporting data for longitudinal research which studies similar problem.

2.      Characteristics

Generally, survey research has characteristics. To be taken into account, the characteristics presented below sometime do not apply to every survey research. Please refer to the research objectives.

a.    Use of large samples

b.   Use of tests, questionnaires, and surveys

c.    Focused on information related to preferences, attitudes, practices, concerns, or interests

d.   Statistical analysis of numerical data

B.     Correlational Research

This is a type of research which is conducted to determine the relationships among two or more variables. Correlational research is an example of what is sometimes called associational research. In associational research, the relationships among two or more variables are studied without any attempt to influence them. Correlational researches investigate the possibility of relationships between only two variables, although investigations of more than two variables are common. In correlational research, there is no manipulation of variables (Fraenkel & Wallen, 2003). Sometimes correlational research is referred to as a form of descriptive research because it describes an existing relationship between variables. Knowing that human behavior is complex, to obtain full understanding studies which investigate the relationships between factors or elements are taken (Cohen, Manion, & Morrison, 2000).

Regarding the presence of two or more variables and the inquiry of relationship, some types of relationship may occur. Sugiyono (2012) classifies three types of relationship, namely symmetry, asymmetry (causal), and interactive (reciprocal).

Symmetry

This type shows relationship between variables which in coincidence emerge simultaneously. Between the variables studied, neither independent nor dependent variable exists. For example, a research question like ‘is there any relationship between hair color and leadership skill?’ would show symmetry relationship. In fact, having certain hair color does not cause someone to be able to lead well.

Asymmetry

This second type shows causal relationship between variables, therefore the variables are independent and dependent. This type of relationship is more familiar in scientific research because it studies how one variable influences others. An example of research question for this type is ‘Is there any relationship between the frequency of listening to English songs and achievement in listening skill?’ The study would show that the frequency of students listening to English songs would influence their achievement in listening skill.

Interactive/Reciprocal

Unlike the asymmetry type, an interactive relationship shows that the variables influence each other, therefore independent and dependent variables cannot be decided. An example like relationship between students’ learning motivation and academic achievement shows an interactive relationship. The students’ learning motivation would influence their academic achievement, or vice versa.

The three types of relationship aforementioned should be put into consideration. Whether variables going to be studied have symmetry, asymmetry, or interactive relationship, so that appropriate research question can be set.

Different from the types of relationship, the following are the types of correlational research. The difference lies on the purposes for conducting the study.

1.      Types of Correlational Research

There are two types of correlational research, namely explanatory and prediction.

a.      Explanatory

This correlational research is conducted when the research problem is to explore the relationship between two or more variables, in a way that changes experienced by one variable may be reflected in the other variable (Creswell, 358:2008). When conducting an explanatory correlational research, researchers typically collect data at one time as their focus is not based on future or past performance of participants. Thus, when analyzing the findings of explanatory correlation research, researchers analyze participants as a single group rather than creating subcategories of participants. Finally, in this type of study researchers collect two scores from each participant as each score represents each variable being studied (Creswell, 2008).

b.      Prediction

It is used when the purpose of the research is to predict certain outcomes in one variable from another variable; one of two variables serves as the predictor. Prediction designs involve two types of variables: a predictor variable and a criterion variable. While the predictor variable is utilized to make a forecast or prediction, the criterion variable is the anticipated outcome that is being predicted. The time at which variables are measured also differs in prediction studies as the predictor variable is typically measured at one time while the criterion variable is usually measured later. Prediction research also includes a forecast of anticipated future performance, as well as advanced statistical procedures including multiple regressions (Creswell, 2008).

After knowing the types of correlational research, it is equally important to understand some patterns of relationship in correlational research. The patterns which will be presented along with figures would clearly show the number of variables present in a study and how they are related.

2.    Patterns of Relationship

Exploring the relationship of variables is the focus in correlational research. Further, the relationships among variables being correlated create patterns which is also named research paradigm.

By determining the research paradigm, a researcher can focus on the patterns emerging between variables, which later would reflect the research problems to be answered, theories used to formulate hypotheses, and statistical analysis technique. Realizing the importance of considering the patterns of relationship, the following are patterns of relationship to know.

a.      Simple Relationship

There are two variables in the study; one independent and one dependent variable showing causal relationship. A research question which employs simple relationship is ‘How is the correlation between learners’ learning style and their vocabulary achievement?’, in which the independent variable (learners’ learning style) influences the dependent variable (vocabulary achievement) in a simple, linear relationship.

b.      Partial/Multiple Relationship

The pattern involves two (partial) or more (multiple) independent (predictor) variables and one dependent (criterion) variable in a study. Whether the predictor variables have relationship with the criterion variable is the research inquiry. Since there are more than one variable as predictor studied to find the relationship with one criterion variable, more complex analysis are employed, refer to Figure 2 for partial and Figure 3 for multiple relationship. It is needed to find the relationship between the predictor variables and a combination of predictor variables to criterion variable.

   
 

In fact, in employing the partial pattern, sometimes a researcher would find that one or more variables as criterion are not suitable, especially because it has a causal effect on the other variable. An example of partial correlation is a study of classroom attention by Lahaderne in Borg and Gall (1983). In the study, found that students who pay attention more would likely obtain a higher level of school achievement than those who do not. Intelligence, however, seems to have relationship with the level of school achievement, too. The students who are more intelligent achieve a higher level of school achievement. In fact, those who are more intelligent would pay attention more in the classroom and achieve higher achievement. Therefore, in analyzing such partial correlation, the intelligence as a variable is removed because it has influence on both, or the variable is held constant (controlled variable) so that the relationship between the students’ attention and their achievement in the classroom would result in more distinct.

c.       Regression

The pattern in this type looks similar to the multiple relationship, in which there are several predictor variables and one criterion variable. Not simply finding the relationship between variables, in regression the purpose is to know how well the predictor variables predict the criterion variable. Assuming that there is no perfect prediction, the regression pattern is used to evaluate the prediction made by certain predictor variables over the criterion variable. Therefore, the analysis in this pattern is more complex because usually the result is expected to be able to predict the future relationship between variables. Figure 4 will show clear pattern of regression.

d.      Path

The pattern shows that between the predictor variables, there is (are) other intervening variables which posited on the path before the criterion variable. In Sugiyono (2012) in a study there are some variables as predictors; students’ social economic status and IQ, and an intervening variable of students’ motivation on the path as another predictor variable, and learning achievement as the criterion variable. In path analysis, as it is different from the multiple relationship, it is analyzed if it is needed for any of the predictor variable to go through the intervening variable so that high correlation with the criterion can be obtained. In order to have better understanding about path, see Figure 5.

 

3.    Data Analysis

A correlation coefficient, symbolized by r, is produced when variables are correlated. The statistic that expresses a correlation statistic as a linear relationship is the product–moment correlation coefficient.

a.      Indicator of Relationship

b.      Meaning of r Values

The coefficient is in decimal and it ranges from +1.00 to -1.00. The variables correlated have stronger relationship if the value is closer to +1.00 or -1.00. If the coefficient is at or near .00, it indicates that there is no relationship between variables involved. The positive sign means if one variable has high scores and the other variable has high scores. It can also be said that the relationship is positive. The negative sign means if one variable has high scores and the other variable has low scores. Therefore, the relationship is negative.

c.       Content of r

Researchers need to be able to interpret what the correlation coefficient means.

Interval of Coefficient	Level of Relationship
0.00 – 0.199	Very low
0.20 – 0.399	Low
0.40 – 0.5999	Neutral
0.60 – 0.799	Strong
0.80 – 1.00	Very strong

The coefficient correlation shows the relation between variables and it applies in the sample used in research. How do we know that the correlation also applies to the population? After knowing the strength of the association between two variables from the coefficient, how do we know if the value is meaningful? One way to find out is to use significance testing. The formula is stated below.

                                                     

The value results from it then is compared with t_table value. If the t_count is smaller than t_table, Ho is accepted and Ha is rejected (the null hypothesis is accepted which means that there is no association or relationship among the scores in the population). If the t_count is bigger than t_table, Ha is accepted (the null hypothesis is rejected which means there is relationship among the scores in the population).

Even though the correlation coefficient is used to determine the strength of relationships, many researchers prefer to square the coefficient and use it as determination of strength in relationship. That calculation is called as coefficient of determination. According to Creswell (2012), “in correlational research, the r squared expresses the magnitude of association between the two variables or sets of scores.”

4.      Types of Correlation

1.      Positive Linear Correlation

This type of correlation is when a scatterplot shows a linear pattern. It means the variable increases when the other variable also increases. When the variable decreases, the other variable will decrease. So, the variables here are related each other significantly.

Example: If students read more, the more knowledge they obtain.

2.      Negative Linear Correlation

It also shows the relationship between the variables, but this type of relationship is negative relationship. It means that when one variable increases, the other variable decreases and opposite from it. Data points are clustered in linear pattern extending from upper left to lower right.

Example: The higher the plain, the lower the temperature.

3.      No Relationship

It means that there is no meaningful, significant relationship between two variables. The points of scatterplot are random position or no pattern. Here, the correlation coefficient for data is very close to 0 (- .09).

Example: A study may find that there is no relationship between students’ intelligence and their weight.

4.      Curvilinear Relationship/ Non-Linear Relationship

If a correlation coefficient of 0 indicates no meaningful relationship between two variables, in this type is still possible for correlation coefficient of 0 to indicate a curvilinear. Both variables still have correlation but in very low correlation. Data points in this type are clustered in a curved linear patter like U shape.

Example: the older the age, the stronger the memory, however, at a certain point, the memory gets weaker even loses.

5.      Characteristics

The correlational research is often confused from the causal-comparative. Therefore, to minimize such confusion, the following characteristics are presented.

a.       Measurement with a correlation coefficient

b.      One group of subjects measured on two variables

c.       Use of instruments to measure variables

d.      Focused on the direction and nature of the relationship