Research Designs

Various research designs can be utilized to study issues of human development as well as other interests. Each design has advantages and disadvantages. It will be important for you to have a general understanding of these designs and how they might be utilized in the study of human development.

Looking at Life Span Development: Independent Variables
Let's get one thing clear right from the start; there are many independent variables that influence human development and human behavior (the outcome). Sometimes it takes a long time for the variables to effect the outcome and sometimes it takes a short amount of time. The variables that influence the behavior outcome are referred to as Independent Variables and the outcome itself is referred to as the Dependent Variable.
For example: we wonder if soda pop increases the likelihood of cavities. Since "cavities" is the outcome we are interested in, "cavities" is the Dependent Variable; and since "soda pop" may influence the outcome, "soda pop" is the Independent Variable.
Are we clear on that now? ....... Good, let's move on.
The "Age" Issue: "age" as an independent variable
Since many characteristics of human development (outcome) emerge slowly over time, it is necessary to consider the time issue or the age issue as an important independent variable in research. This means that some outcomes are a result of the "aging process" and gradually change as time passes. Therefore, some research may take considerable time to complete. Since we have a life span of approximately 85 years, some research may go on for many years. If we decide to study the entire life span of humans as they grow from infancy to old age, we will need to use a design that will allow us to complete the study within our own life time. Some of the designs discussed here will allow us to study many years of development in a relatively short period of time, while others will require longer periods of time to complete the study.
* Since life span research looks at how people grow and change over time, as a result of time, the issue of "age" is of major importance as an independent variable. Biological influences gradually unfold over the life span and so we are often interested in how "age" is a part of this biological process.
Other Issues: other independent variables
However
... there are other issues and questions in development that are NOT AGE RELATED. These are other independent variables we are interested in. There are variables such as socioeconomic status (SES), level of education, religion, gender, drug use, abuse, exposure to chemicals, diet, etc. There are many variables that influence development of the person besides age or aging. We are interested in all of the variables (independent variables) that influence the outcome of the person (dependent variable). We will discuss this more in a few moments.

Representative sample:
Since we cannot study the entire human population, (way too big of a sample) we take a sample of the population that we believe represents the larger population. This small sample is referred to as a "Representative Sample". This allows us to look at certain issues of interest in development (independent variables) and then generalize them to the larger population. It is very important to keep in mind that this sample must truly be a representative sample of the population we hope to generalize about. In fact, we will only be able to generalize results to this sample population and a larger population that it represents. In other words, if we use a sample of 16 year-olds in Utah to study drug use, then we will only be able to generalize the results to other 16 year-old in Utah, and only to those 16 year-olds who live in the same general area as our sample population.

An Example Research Question:
If we are interested in cognitive development and the changes that take place between the ages of 5 years and 20 years, how will we study this issue? Note that this is an "age" related question. "Age" will be the independent variable in this case and "cognitive development" will be the outcome or dependent variable.


Cross-sectional: When studying humans, this design uses a representative sample of a larger population, with the hope of generalizing the findings to the larger population. Collect data once on the variables of interest and look at the correlation between the variables. Do not mess with the sample population. Do not alter the independent variable for this population. The cross section or "slice" is across the independent variable of interest. The independent variable might be age, or some other variable you are studying, measuring, looking at, or interested in. NOTE: the cross section is always across the independent variable.

This will usually allow for generalization of the findings to the larger population, as long as the sample is truly a representative sample. Because we want to get this research done in a short period of time, we will use the cross-sectional design. We will want to select a sample to represent the various ages in development at ages 5, 10, 15 and 20 years (they are alive right now, we will use them and collect data from them now in time). This sample could be considered a "slice" of the general population of people living right now within those age groups. The "slice" is across the age variable. Age would be an independent variable. The researchers will collect data on these people and assume that they represent typical development of others who are also that age (5, 10, 15, & 20), or who will be that age at some time. Data is typically collected at one point in time (now) on the research sample. This allows for data to be analyzed and findings to be published in a short period of time (days or weeks).

The assumption is that the 5 year olds in the study will be like the 20 year olds in another 15 years.... and that the 20 year olds were like the 5 year olds 15 years ago. Well, because there are social and historical issues that can effect development of a particular birth cohort group, this assumption is not always correct. This has reference to cohort effects in development, the possibility that social, historical, economic events can influence and shape the attitudes and behaviors of a particular generation (birth cohort). For example: did the depression of the 1930's (economic event) have an impact on that generation of young children?
NOTE: Cross-sectional research does not look at cohort effects.

The researcher collects data from the sample on the independent variable(s) and the dependent variable and looks at the correlation to determine what the relationship is, if any, between these variables. See information on correlation to better understand how this works.
NOTE: The "slice" in cross sectional research DOES NOT have to be across age. It is across the independent variable being studied. If "age" is the issue in question, then the slice is across age. If "education" is the issue of question, then the slice is across level of education.
Do not mess with the sample population, just collect the data and look at the correlation.


Longitudinal: Another method of looking at development over time is the longitudinal study, following the same group of people over SOME period of time. Collect data more than once on the sample population over SOME period of time. Look at the correlation between the variables of interest. Do not mess with the independent variable for this population, just follow their development and collect the data and look at the correlation.

This design also uses a representative sample population, so generalizations can be made. However, this population will be studied over time as they develop and data will be collected on this sample more than once during that time. If we use that same age span 5 years to 20 years, this would mean that a sample of 5 year olds (birth cohort group 2005) would be selected and data would be collected when the population was 5 years old, 10 years old, 15 years old, and 20 years old. Obviously, this will take 15 years to conduct the study as opposed to a few weeks to collect data with the Cross-sectional method. Because time is an issue, some members of the sample may drop out, move away, or even die before the study is completed. This type of research can be expensive as well as time consuming. It does allow the researchers to assess the actual changes in development as they are observing the same people over time. AGAIN: You do not mess with the sample population. Just observe and collect data and look at the correlation to see what the relationship is between the variables.

Cognitive development is a slow gradual process of change. You will need to observe this process over a few years to get a good perspective of it. If the issue of interest (dependent variable) is something that changes rather quickly (like a scratch on the arm), then a lot of time is NOT NECESSARY. A shorter period of time would reveal results that could be generalized. So... you may not need a long period of time for this design if you are studying something that could change in a short period of time. You just collect data on the variables of interest, wait some time, collect data again from the same people and look for significant change in the dependent variable.

Consider the question of which Band-Aid helps heal a scratched arm more quickly.

The researcher collects data from the cohort sample on the independent variable(Band-Aid) and the dependent variable (scratch on the arm) at various times during the study. At the conclusion, the researcher looks at the correlation to determine what the relationship is, if any, between these variables. Various types of Band-Aid are already used by this population so you will not need to assign people to a particular "Band-Aid group". We hope to determine if one Band-Aid is better than the others. Since a scratch will heal in about 5 days, you will not need a long time for this study. See information on correlation to better understand how this works.

It is important to note that the researcher DOES NOT alter the independent variable in a longitudinal study. Data is collected on the variables from time to time, the sample goes about life as usual, and comparisons are made concerning the dependent variable as the individuals age. Remember: DO NOT MESS WITH THE INDEPENDENT VARIABLE.

Note the designs below. When "age" is the independent variable:
Cross-sectional: population made up of different cohorts (1980, '75, '70, '65) and studied at one point in time (1985). They represent people of ages 5, 10, 15, & 20 years, who are that age right now. Collect the data and look at the correlation to make a determination.
Longitudinal: population made up of cohort 1980 and studied in 1985, '90, '95, & 2000 when they are 5, 10, 15, & 20 years old. You follow the same people for those 15 years and collect data along the way. You can see how that would be expensive and time consuming for the researcher. Collect the data and look at the correlation to make a determination.


Now, let's look at the Experimental Design

First... a little review about variables.

Dependent Variable... this is always the outcome variable you are looking at and measuring. The outcome always depends on other variables (independent variables). If someone asked you "How do think your children will turn out as they grow up?" You might say.... "Well, that outcome depends." Of course it depends. The outcome always depends on all of the other variables that might effect the outcome. So, the "outcome variable" is always the "dependent variable", since it depends on other impacting variables.
Independent Variable... this is the variable that is believed to have some impact or effect on the outcome variable. This is the variable that you mess with when doing an experiment, such as changing the amount of TV violence children are exposed to and watching for changes in aggressive behavior, or altering the background music in a store to see the effect on shopping behavior. The independent variable is the variable you think has some effect on the outcome (dependent variable).

Experimental: This method also utilizes aspects of the longitudinal design since in most cases researchers are examining the effects of a particular independent variable as time passes by manipulating the independent variable for some of the sample population. By controlling the variables it allows the researcher to infer a "cause and effect" relationship between the variables.

Representative Sample:
As with other designs, this design also uses a representative sample population, so generalizations can be made. You start with a large population of people who are very much alike in many ways. You want them as alike as possible in your attempt to control as many independent variables as possible. Once you have your initial population, you randomly assign participants to various groups for the study. The groups will be described below.

Data is collected on the sample at various times during the study. This method of study is a controlled study where researchers hold most variables constant while manipulating a particular independent variable. Since the purpose of this design is to address the issue of "cause and effect", only one variable is manipulated at a time. You should be able to see the problem you would have if you were manipulating 2 or 3 variables at a time. That is correct.... you wouldn't know which one was actually causing the effect if you were messing with 2 or 3 independent variables at the same time.

How about the sample... the people in the experiment?
Remember... you started with a population of people that are very much alike. You then randomly assign them to various groups for the study.

Experimental Group... this is the group of participants that you are messing with... you adjust or change what they are exposed to. Often this is referred to as the "Treatment Group" because they get treated differently than their usual behavior. With this group, you alter the independent variable (increase it or decrease it or introduce a different form of it... ). In other words, they get treated differently than they were used to before the experiment. There can be more than one experimental group you are looking at, each with a different level of the independent variable you are manipulating. If you are studying the effects of chocolate 1 group might double their chocolate and another group might triple their chocolate and another group might cut their chocolate in half. Remember that all participants ate about the same amount of chocolate when we started.

Control Group... this is the group of participants that you DO NOT MESS WITH. This is the "life as usual" group. They do not get special treatment, they go about their life as usual. You collect data on them over time (like a longitudinal study) so that you can compare them with the Experimental group. The reason for this is that some things may just change naturally over time (due to maturation), and not because of any treatment. The control group lets you note these "natural changes" and compare the treatment group(s) with a "life as usual" group to see if there really are any significant differences between these groups when looking at the dependent variable (the outcome). "Significant" means statistically significant, and not by chance. In the chocolate example, these people would continue their life by continuing to eat the same amount of chocolate as usual.


Why is a control group needed?
If we consider reading ability as a dependent variable (outcome), we will have to take into consideration that reading ability will naturally change over time, as a result of growth and not due to any special treatment. By using a control group we can observe the changes as effected by time and then compare the two groups to see if there are significant differences.
Since both groups experience the passing of time, any statistically significant differences in reading ability are assumed to be a result of the "treatment" or independent variable received by the experimental group.
If there are no statistically significant differences, but both groups increase in reading ability, then we would conclude that the increase was due to "aging" or the passing of time.
Without a control group, you are not able to make any comparisons.


Now... an example of an Experimental Design: Music and Shopping behavior
A grocery store owner wonders if background music influences shopping behavior. He wonders if a certain type of music will keep shoppers in the store longer, because he knows that the longer one is in the store, the more money they spend. He currently DOES NOT PLAY MUSIC in any of his 5 stores.
Note: Assume that this is when background music was not used in stores (before your time) , so the norm was No Background Music. In other words, "No Music" was life as usual for shoppers..
We want to know if there is a "cause and effect" relationship, so we will set up an experiment to look at the relationship of these 2 variables. Consider the following information and see if you can fill in the blanks.

Independent variable: ___________________________________________________
Dependent variable: ____________________________________________________
Experimental groups: ___________________________________________________
Control group: ________________________________________________________

 

We are not only looking at the presence of background music, we are looking at "type" of background music in this example. Remember... since the norm has been "no background music" any alteration of this would constitute a modification of the independent variable, and would make that situation an "experimental group". The "life as usual" group (no modification) would make that the "control group".

The study would be set up this way. Start with a sample of shoppers who are very much alike and have not had background music while shopping. They are randomly assigned to the various groups. There will be 2 experimental groups (A. Classical music and B. Rock & Roll music) and a Control group (C. No music as usual). Since the effect of music is in question, it is the independent variable. Since shopping behavior is the outcome being measured, it is the dependent variable. In one the stores the owner pipes in Classical music, in another store he pipes in Rock & Roll music and in a 3rd store he leaves it as it is, without any music (control group). They have noted the total sales for each store before the experiment for a baseline comparison. They collect sales data over the next 3 weeks and compare the results. The owner wants to know if there is any significant difference in total sales (shopping behavior) between the experimental groups and the control group. A correlation of the variables will give the researchers an idea of what to conclude.

 

 


Control Group is:
<--- Life as usual

No modification of Independent variable for the control group..

Remember.... the Control Group is the "No Special Treatment" group, don't mess with these people. That means that they are not exposed to the treatment (do not alter their live style) and they go about their lives as usual... shopping without music. You do not alter the Independent variable for the Control group, GOT THAT?

If we find that background music does effect shopping behavior by causing them to spend more time in the store and spend more money, we will want to play the type of music that causes people to stay in the store longer and spend more money.

This design gets to the issue of "cause and effect" where other designs are looking basically at the correlation between the variables. As you know, correlation is NOT causation.


And now... the correlation:

Correlational:

This method is used to study the relationship between variables
For additional information Click HERE

Sometimes this method is used by utilizing data that already exists. For example, using census data as the data set, a researcher might want to look at the relationship between family income and level of education. The data is already there waiting to be used. The researcher could pull the data on these variables and calculate a correlation to look at the relationship that education might have with income. In other words, you might wonder if people with more education make more money (income) than people with less education. The correlation will tell us what the tendency is; as education increases, does income tend to go up or down or stay about the same??

A correlation is also valuable when combined with the other designs of research. When doing cross-sectional research, a correlation coefficient will be calculated to give the researcher a picture of how the variables are related. As variable X increases, what does variable Y do? Does it also go up (positive (+) correlation) or does it go down (negative (-) correlation) ?

You should note: ... this is only a description of the relationship between the independent and the dependent variable. This is NOT an indication of cause and effect. Even if the correlation is strong (.95), it only describes the relationship between the variables, it does not imply that variable X is causing variable Y to behave that way. There could be a third variable that is influencing both X and Y to behave this way in relation to each other.

For additional information on correlation (and scatter plots) see the link on the Learning Objectives page....
or Click HERE


Lastly, you might consider the Cross-sequential design:

Cross-sequential: (sometimes called Time-lag sequential or Cohort sequential)

This design utilizes a combination of longitudinal and cross-sectional methods. It actually is a series of overlapping longitudinal studies, started at different times and using different cohort groups.

Refer to the chart below:
If we are studying the various age groups mentioned above (5 yrs - 20 yrs), you would select a population from the 1970 birth cohort (brown boxes below) and study them in 1975 when they are 5 years old and study them again in 1980 when they are 10 years old. Also in 1980 you would pick up a new sample from the 1975 birth cohort (green boxes below) and study them while they are 5 years old.

Then, you will notice, in 1985 you would study the 1970 birth cohort (brown boxes now 15 years old) and the 1975 birth cohort (green boxes now 10 years old) and you would start a new 1980 birth cohort (white boxes now 5 years old). This allows you to make comparisons across cohorts, comparing 5-year-olds of different cohorts, 10-year-olds of different cohorts, and so on. Each birth cohort represents a longitudinal study as you study them over time. The chart below shows how these cohorts overlap.

One of the reasons this design is used is to allow the researchers to look at cohort effects, since different cohort groups are represented by the overlapping longitudinal samples. Remember: cross-sectional does not look at cohort effects. When "age" is the indpendent variable, it picks participants from different cohorts who are alive now to represent those particular cohort groups. It does not study them over time.

Because this design is so complex, it is even more expensive and time consuming than the longitudinal design. Although it can be a very effective way of collecting some very usable data, the limitation of money and time makes this design seldom used.

If you really wanted to get a research study done in a hurry and get some results published, you would stick with the cross-sectional design.

For additional information on correlations Click HERE