Statistics and Lies: Assessing the Value of Collected Data.
Researchers and those who work from knowledge based upon their research findings rely heavily on statistical reports that are constructed in many different forms. There are two main branches of statistical methods - descriptive statistics which behavioral scientists use to summarize and make understandable, or to describe, a group of numbers from a research study (Aron, Aron, & Coups, 2008).
The other main branch of statistical methods is inferential statistics. Behavioral and social scientists use inferential statistics to draw conclusions and inferences that are based on the numbers from a research study, but go beyond these numbers (Aron et al, 2008).
As stated above, scientists collect data and create what we call bars and graphs from the information gathered in order to make sense of patterns of behavior or other kinds of data. When their information is translated into a display using tables or graphs, the information is easily interpreted, in most cases, and depended upon to support theories and beliefs about the particular study being done.
For example, a frequency table shows the approximate number of times a certain behavior occurs. In a study that investigates the amount of stress students are experiencing, the variable (stress) may be determined using a scale of 0 - 10, perhaps. 0 would represent the least amount of stress possible, and 10 would represent the most amount of stress a student could feel.
Now let's say this study is being done by a university professor who teaches this class, and who does not want to indicate that his class is particularly stressful. He may do a number of things to impress upon the students beforehand, the importance of positive responses, or try to influence the students' responses this way. Or he may construct the questions in a double negative or other format that purposely confuses the students into responding with only answers that reflect positively on the professor. There are a variety of ways the study can be skewed to produce a good outcome. The more students reporting small amounts of stress would show that the class is not stressful, and that the professor has provided a relaxing and welcoming atmosphere in the classroom, making the learning process as stress-free as possible. He or she has a personal interest in the study.
Conversely, the same professor may want to impress upon the university that the material he or she is teaching is much too difficult for his freshman class. A study he creates may ask questions that will skew the results to produce statistics that show great amounts of stress attributed to the subject material. If 85% of the students rate their level of stress in the classroom at 8 or 9 out of 10, the instructor would have an easier time proving his case. Again, he has an interest in the study, and can skew the results in favor of his argument.
Now suppose the students themselves do not wish to be seen as stressed or unable to handle the subject material. It is possible that they would misrepresent their answers and say they have only minimal stress in the classroom, perhaps rating themselves at 2 or 3. They may actually be experiencing far greater levels of stress, but are unwilling to share this on the study for fear of reprisals from the professor or of being seen as incapable of handling difficult learning concepts.
In any of these statistical studies, the value of the collected data is practically nil. When the results have been skewed in favor of one position or another, or the respondents have misrepresented their answers, the outcome is highly undependable. In the current political atmosphere, statistics are used all the time to determine what percentage of voters lean toward particular candidates, and those results are broken down into groups of voters, and the groups are broken down into smaller groups, and so forth. Depending upon the political slant of the researchers, it is possible that the statistics presented here are inaccurate in order to persuade voters to follow what appears to be the popular trend for their group.
The bottom line is that statistics can be an amazing tool to help scientists or researchers of all kinds understand patterns in the subjects they are studying. The information that is gathered can be genuinely valuable and can help predict future behaviors or events. But it is possible that the value of collected data can be reduced or eliminated entirely if the purpose of the study is self-serving. One must consider the source of the collected data to determine if it has significant value.
Aron, A., Aron, E., & Coups, E. (2008). Statistics for the behavioral and social sciences: A brief course. (4th ed.). NJ: Pearson Prentice-Hall.