We have dealt with ways of "examining" data mathematically. But how is data collected? Is there a "best" way to obtain data? Statistics is concerned with the collection and analysis of data. There are several different "types" of statistical studies that are used to collect data. Different study types have their advantages and disadvantages. Let's take a look at the two major types of statistical studies: experimental studies, and observational studies.
Review the Four-Step Process for conducting a statistical study and avoiding bias.

Collecting Data:
There are a few standard ways of collecting data: use already existing data, survey a sampling of the population, design an experiment, use a census, and use a simulation.
• In relation to statistics, a "census" is broadly defined as a list of ALL individuals in the population along with characteristics regarding each individual. A "census" is extremely difficult to obtain. Imagine trying to survey millions of college students, for example, to determine if college students preferred iPhones or Android phones.
• A "
simulation" uses a model (such as a computer model) to replicate the conditions of a process or situation. Simulations are commonly used when actual conditions are too expensive, dangerous, impractical or unethical to replicate in real life.

Observational Study -  In an observational study, the sample population being studied is measured, or surveyed, as it is.  The researcher observes the subjects and measures variables, but does not influence the population in any way or attempt to intervene in the study.   There is no manipulation by the researcher.  Instead, data is simply gathered and correlations are investigated. Since observational studies do not control any variable, the results can only allow the researcher to claim association, not causation (not a cause-and-effect conclusion).

 An example of an observational study: Is there a correlation between attending an SAT Prep class and scores achieved on the SAT Examination for this school year?  In an attempt to investigate this possible correlation, a group of students who took the SAT Examination are surveyed. The scores from students who took an SAT Prep class are compared with the scores of those that did not take an SAT Prep class.  A statistical analysis is performed on the data. This is an observational study since the researcher did not manipulate the sample set.

Surveys - Surveys are one form of an observational study, since the researchers do not influence the outcomes. Statistical surveys collect information from a sample group to learn about the entire population.  A survey may focus on opinions or factual information depending upon the purpose of the study. Surveys may involve answering a questionnaire or being interviewed by a researcher.  The U.S. Census is a type of survey.

 Advantages of surveys: Disadvantages of surveys: • can be administered in a variety of forms (telephone, mail, on-line, mall interview, etc.) • are efficient for collecting data from a large population • can be designed to focus only on the needed response questions • are applicable to a wide range of topics • are dependent upon the respondent's honesty and motivation when answering • can be flawed by non-response • can possess questions or answer choices that may be interpreted differently by different respondents (such as the choice "agree slightly") Randomization and a well-designed survey: A sample population is considered random if the probability of selecting the sample is the same as the probability of selecting every other sample.  When a sample is not random, a bias is introduced which may influence the study in favor of one outcome over other outcomes.

Designed Experimental Study
- Unlike an observational study, an experimental study has the researcher purposely attempting to influence the results. The goal is to determine what effect a particular treatment has on the outcome. Researchers take measurements or surveys of the sample population.  The researchers then manipulate the sample population in some manner.  After the manipulation, the researchers re-measure, or re-survey, using the same procedures to determine if the manipulation possibly changed the measurements. Since variables are controlled in a designed experiment, the results allow the researcher to claim causation (a cause-and-effect conclusion).

In a randomized experiment, researchers control the manipulation of the sample populations using a chance mechanism, such as flipping a coin or using a computer to generate random numbers. Randomization is important for experimental studies so the researcher can know that it was the treatment given to the population that caused a change in the population.

During a "controlled" experiment, the researcher will separate the sample population into groups with one group established as the control group.  All groups will be manipulated in some manner, except for the control group which will remain the same.

 An example of an experimental study: Does the color of a basketball influence the number of times a shooter sinks a basket? A random group of students is chosen and asked to shoot a series of baskets using a regulation normal-colored basketball.  The data is recorded. The same group is then given a blue colored basketball and the same number of shots is repeated. The data is again recorded. A statistical analysis is performed. This is a designed experimental study since the researcher manipulated the conditions of the study by changing the color of the ball.

Comparison of Observational Study and Designed Experimental Study:

 Observational Study Designed Experiment • Observe only, no "treatment" assigned • Generally a control group is not needed • Reports an association • May (or not) use random sample sets • May (or not) generalize to population • "Treatment" assigned • Uses control group for comparison • Reports a cause and effect • Randomization of sample group • Generalize to population