This page goes over the definitions of some statistical terms.
A source is a place from which someone gets information. Claims need sources to back them up, unless the claim is common knowledge. If someone claims that they saw a Ferrari driving down a street in town, they can back it up in at least a few ways, including a picture, footage, or the confirmation from several other, trustworthy people. In the same similar fashion, if someone claims that a harm-reduction stance toward drug addiction is better for society than an abstinence-only stance, they better have an arsenal of expert opinions and statistically rigorous studies to back up that claim.
But sources can be dead wrong. Studies must be done in a particular manner to have validity, Principles of Qualified Sources, in order to be taken seriously. I’m listing two principles here, which can also be found in the principles section, because they come up in every page that I cite studies:
- Meta-analysis: this study form looks at many standalone studies that shared a research question, then though strict criteria uses the ones that have a similar enough design and method. The researchers design statistical equations based on the results of these standalone studies, and run them to reach a quantitative conclusion.
- Systematic review: takes into account all of the studies that have been conducted on a topic. It’s more general than meta-analyses, and isn’t run through as many tests.
Any scientific exploration begins with a hypothesis: an educated guess regarding the relationship between different things. A hypothesis becomes a theory after a significant amount of testing affirms the hypothesis.
Statistically rigorous principles are either present, or absent, in the full-text: the entire layout of how the study was done.
Here we have the general structure of a full-text…
- Title and authors
- Affiliations
- Perhaps of a certain hospital or university
- Date of study
- Within ten years is generally considered recent enough
- If it was peer-reviewed, and if so, when
- Essential
- In which journal it was published
- This matters, because journals vary in prestige; The Lancet and Nature are highly prestigious
- Affiliations
- Abstract
- At the top of a study
- A paragraph, 250 words or so, summing up a study
- Includes
- Purpose
- Design specifics
- Findings
- Interpretation of results
- Table of contents
- Introduction
- Overview of prior, relevant research
- Describes bigger picture
- States a specific research question, and how it fits into a larger framework
- Methods and materials
- Statistical techniques and tests
- For instance, a random number generator isolates a representative sample from a large population of like objects or participants
- Describes what was done clearly enough that it can be replicated
- Statistical techniques and tests
- Results
- Determining which findings are significant
- Discussion
- Findings are analyzed
- Findings, and pre-study hypothesis, are compared
- How the findings fit in with the larger picture, or more general subject
- Conclusions
- A more detailed version of the abstract
- More opinionated than the abstract
- References
- Acknowledgements
- Thanks people who helped indirectly
- Lists those whom helped fund the study
A variable is something that’s being measured. For instance, one’s height.
- Independent variable
- The variable controlled by the researchers
- For instance, the amount of milk given to children in a study on bone health
- Dependent variable
- A variable that’s measured after the introduction of the independent variable
- For instance, the health of bones after varying levels of milk
- Extraneous variable
- In an experiment, variables that aren’t being accounted for
- Can turn into confounding variables, and sabotage the experiment
- Confounding variable
- An extraneous variable that is significantly associated with the independent variable, the dependent variable, or both the independent and the dependent variable
- Sabotages the study
Next we have coorelational research, a which roughly describes the relationship of two variables. Coorrelational studies are simple. They can never determine causation, as experiments can. The two variables in question can either have a negative relationship (as one increases, the other decreases), or they can have a positive relationship (as one increases, the other increases)
- Coorelation coefficient: A number, from -1.00 to 1.00, that describes the strength of relation of two variables
- Known as R
- A negative number means a negative relationship, a positive indicating a positive one
- The closer this value is to 1.00, the more strong a positive relationship
- The closer this value is to -1.00, the more strong a negative relationship
- Coefficient of determination: a number that bears on the accuracy of a model, how much variation there is in an independent variable relating to a dependent variable
- Known as R^{2}
- Outlier: a data point or value that significantly deviates from the general trend
- Experiment
- A controlled study that measures the effect of an independent variable on a dependent variable
- Extraneous variables are minimized
- Can much better determine causation than a correlational study
- Randomization controlled trial (RCT) of which objects or participants are given the intervention, and which are given the placebo
- Experimental group
- Given the intervention, an active treatment
- Control group
- Given the placebo, which appears exactly as the intervention, but is fake
- Data
- Facts derived from study
- Graph
- A composition data, can take many forms
- Mean
- The calculated, numerical average
- First instance, 5+4+12+16=37. 37/4=9.25. So the average is 9.25
- Median
- Middle value
- Mode
- The most common result
- For instance, a study involving the age of onset for schizophrenia would find the mode to be 18, the most frequent age at which the disease began
- N
- Number
- Denotes how many participants, or objects, were included
- For example, N=34 means that 34 were included
- Applied research
- Research that is directed toward some real-world goal, to solve a problem or make something easier or better
- Example: testing out a chemical to see if it’s effective at reducing clinical depression in minutes or hours, whereas the antidepressants available take weeks and possibly months to work entirely
- Pure research
- Research out of bare curiosity, just to learn
- Example: tracing back family, finding out ancestors and creating a family tree
- The information isn’t to make things better. Knowledge itself is valuable
Sources: Dr. Laura Wray-Lake, Ben Komor, Karen Siefert,