Avoid Biased Data It is a goal of every reliable statistical study (especially a "sampling") to avoid being labeled as a "biased" study. A biased sample is one that is not representative of the population from which it is drawn.

Statistical bias happens when "favoritism" in the data collection process
(or the reporting process) occurs, resulting in misleading results.

When dealing with data, it is possible that different statistical studies,
concerning the same issue, can arrive at very different results.

For example, one study shows that students who reviewed for the SAT examination for five or more hours, scored in the top 10% of the students taking the test in the spring of 2023. A second study, also relating to the spring of 2023, showed that reviewing for the SAT examination did not result in scores in the top 10% . How is this possible? Shouldn't statistics always arrive at the same results concerning the same issue? Not necessarily.

The manner in which statistics are reported may accidentally (or intentionally) support a specific desired result. For example, a drug company publishes in a magazine a study that found positive results from the use of their knee-joint supplement, but does not publish a study that found negative, or even dangerous, results from use of the same supplement. Consumer beware!

Due to these influencing factors, it is important to understand how to avoid bias situations when using statistical data in your research, as well as how to recognize research which may be exhibiting bias.

Here are a few of the questions to ask yourself when dealing with data:

1. Who is collecting the data?
Does the group collecting the data have an interest in the final results?
For example, a drug company funding research on the safety of their product may result in unreliable findings, due to a conflict of interest. The research should have been carried out by a third party that was not connected to, or paid by, the drug company.

2. How was the data collected?
Was the sample (the group used for the study) truly representative of the population (the much larger group to whom the findings are directed)? Was the sample group chosen at random?
Statistical bias can be avoided by using random statistical samples.
For example, 300 people, upon exiting the last Twilight Saga movie, were asked to respond to a survey regarding their preferences regarding types of literature enjoyed by Americans. Only 50 people completed the survey. Since only 50 of the 300 people responded, this was an insufficient number to be a representative sample. In addition, since it may be the case that the majority of the Twilight Saga movie-goers were teenagers, the sample may have already been biased toward a particular genre of literature.

3. Were reliable measuring instruments used?
If measuring devices were used in the collection of the data, were the devices reliable, accurate and appropriate for the study?
For example, a digital human weight scale may not be sufficiently accurate for samples of extremely small size such as grams of sodium, thus yielding unreliable data for certain items in the sample.

4. What was the sample size of the study?
How many people, or items were studied? Were there a sufficient number of people (items) in the study to generalize the findings to the entire targeted population?
For example, a survey asks high school students in French club, if they drink black coffee. Since there are only 15 students in French club, there are too few participants to generalize the survey's findings to all high school students.

5. When did the study take place?
Is the study (and the data) current, or did it occur decades ago? It may be the case, that developments since the date of the survey will contradict its findings.
For example, decades ago, it was felt that second-hand cigarette smoke was not harmful. Further research concluded that this was not true, and that second-hand cigarette smoke contains cancer causing agents and is unsafe. Several laws now restrict smoking in public places.