Statistical Testing in Scientific Research
Methods Used to Analyze Data Gathered Through Experimentation
Jul 13, 2009
Statistical methods can be used to summarize or describe a collection of data; this is called descriptive statistics. Such statistics are useful in research, when communicating the results of experiments. In addition, patterns in the data may be modeled in a way that accounts for randomness and uncertainty in the observations, and are then used to draw inferences about the process or population being studied; this is called inferential statistics.
Inference is a vital element of scientific advance, since it provides a prediction (based in data) for where a theory logically leads. To further prove the guiding theory, these predictions are tested as part of the scientific method. If the inference holds true, then the descriptive statistics of the new data increase the soundness of that hypothesis. Descriptive statistics and inferential statistics (a.k.a., predictive statistics) together comprise applied statistics.
How are Statistics Applied to Science?
What is the best and proper use of statistics in science? Statistical procedures are used to determine if a hypothesis is supported or rejected by the experimental data. The appropriate statistical test properly applied can verify whether the data show any significant difference between a control group and an experimental (test) group in a given experiment or if such difference could be attributed to mere chance.
What kinds of statistical tests do scientists use and in what experimental situations are each type of test used appropriately?
The Chi square Test (or goodness of fit test) is used to determine whether or not two or more samples (of seeds, bacteria, or humans) are significantly different enough from each other in some characteristic, trait, or behavior that it is appropriate to generalize that the populations from which the samples were taken are also significantly different. Chi square tests are most often applied to data generated in genetics experiments or situations where there may be two possible outcomes. For example, what feather color(s) are dominant in a wild population of certain birds? The Chi square test could be used to answer that question.
The t-Test assesses whether the means (simple average) of two groups are significantly different from each other in some characteristic or trait. For example, in a rat growth experiment testing a synthetic growth hormone, is the growth rate of the rats fed synthetic hormone greater than that of rats fed no hormone? The t-Test could be used to answer this question.
ANOVA (or Analysis of Variance) is a test applied to compare samples among more than two and up to any manageable number of groups. For example, a researcher wishes to determine the best concentration of synthetic hormone to be dispensed to rats to generate the fastest rate of growth. The rats would be divided into three equal groups – low hormone concentration, medium hormone concentration, and high hormone concentration. The total weight gain over a select period of time of the rats from each group is determined and recorded. The ANOVA test could verify which concentration of hormone, if any, yielded a significantly faster growth rate. ANOVA essentially tests whether there is greater variance among than within groups than would be expected by chance, but it does not tell which particular samples are different from each other.
Statistics Can be Manipulated
Many people have the mistaken idea that “With statistics you can prove anything.” This is incorrect. Statistics cannot prove anything, but if used properly, statistics can indeed demonstrate certain degrees of relationship and significance among factors or variables.
No discipline has been so misused as statistics.
(John Fennick)
Statistics can be manipulated and connections made that are artificial if not outright ridiculous at best. Is living near the equator is dangerous to your health? Studies show that on average, people live about a half year longer for each 100 miles of resident distance from the equator. People who live 1,000 miles north of wherever the reader is located have a life expectancy five years greater than the reader. That is equal to what has supposedly been gained from the last thirty years of medical progress.
Readers may scoff at the artificial contrivance and connection made in the equator case but the evidence supporting these contentions is as good (and in some cases better) than that supporting many statistical “truths” that are sustained with billions of dollars of grant money each year. For example, it is better than the evidence that low-fat diets and exercise have reduced deadly heart attacks. Studies proclaiming doom or glory bombard the public one after the other with the last often contradicting the previous.
What choices should people make, then? What lifestyle changes should people consider to improve and safeguard their health? At present there is no clear-cut answer or easy choice. It remains a matter of asking the right questions and understanding the conclusions.
 |
