This page includes information that builds, and overlaps with, previous material. It will help one navigate through studies if they choose to do their own research.

If they have decided to look up clinical studies, then this guide will clear up some of the jargon.

As far as learning proper statistical method, the best method would be taking a college course in statistics, or the equivalent of one.

*Coursera *is an online institution that contracts out courses from colleges around the world, including Ivy League schools. The subject material ranges the whole of academia. Some courses are specific, others more general.

Usually, one can audit a course for free. Otherwise, it is usually $49 for taking the course for a certificate if one gets a minimum grade.

- These classes that bare on statistics, from
*Coursera,*teach statistics far better than I can:

I took Advanced Placement Statistics in highschool. We ended up learning how to crunch very advanced equations with scientific calculators. This section does not include such knowledge.

Below are some of the definitions and explanations associated with scientifically founded statistical rigor!

**Correlation studies**are perhaps the most simplistic form of study. They may be a good initial study.- The function is fairly simple. A bunch of data points are collected, and plotted on a graph. A line-of-best-fit is then drawn across the graph, and a value calculated.
- The correlation is the nature of the association of two variables. For instance, there is a high association of methamphetamine abuse with brain loss.
- This association is not just high, it is positive. This is to say that as one variable (methamphetamine abuse) increases, so does the other (brain loss).
- An association can still be high and negative. For instance, the variable of years of smoking was found to increase as incidence of lung cancer increased. The variable generated from the plot shows a preliminary message. The “Strength of correlation” depiction below depicts the numerical value that would be generated given specific lines of best fit.
- Importantly,
**coorelation does not mean causation!**Even if the variable is 1, or -1, causation cannot be determined in this form of research.

- Above we see a description of correlations
- This “shot-gun” approach of this method reveals variables that may be related
- The study can be conducted with haste
- This form of study cannot determine causation
- It requires a large sample
- Some correlations are significant by chance
- Causality could be due to a third variable that’s unmeasured
- These are called extraneous variables
- When it is determined that this other variable is more at play, it is called a confounding variable

**Experiments**have high statistical value, in general- They are expensive, though
- There is a control group and as experimental group
- Intervention/quasi-experiment
- This kind of study can determine cause/effect information
- This is due to control
- One, and only one, variable is carefully altered to see the effect on another variable
- The independent variable is the variable altered
- The dependent variable is the variables witnessed and measured after alteration in the independent variable

- cant manipulate everything, poor retest reliability

**Surveys**- Quick and easy
- Can be given to people whom are related to the people under study
- For example…
- Peers of the population under study
- Parents of the population under study
- Teachers of the population under study

- For example…
- Yet can easily be falsified, as their is no overview of the questions being answered
- Wide scale
- Sensitive information
- Lacks social desirability

**Interviews**- A set collection of questions
- Structured
- A checklist of questions to be asked

- Unstructured
- More of a dialogue, with the question being answered therein
- Less formal atmosphere

- Detail, tailoring and probing
- Lacks social desirability
**Case study**- One person is intensively interviewed or observed

**Ethnography**- A description of the population under study
- Meant to be objective
- Participant observation, or, embedded research
- Researcher(s) goes under cover to live among the population under study

- Intense, rich description
- Generalization, researcher bias
- Psychophysiological

**Observational**:- Naturalistic
- More generalization
- Less participant effort
- More data from each subject
- Being familiar with the subjects
- Less control
- Some infrequent behaviors,
- Small sample
- Bias observer
- Usually a parent

- Influence of being observed

- Naturalistic

**Randomization**- Quasi-experiment
*aren’t randomized* - Choosing participants, and treatments in a study, randomly
- Such as by assigning a number to each in the population
- Then, using a random number generator, picking out a sample to study

- Such as by assigning a number to each in the population
- The sample must be representative of the population as a whole

- Quasi-experiment

**Randomized controlled trials (RCT)**- Participants are chosen at random to be part of..
- Experimental group
- The intervention is received

- Control group
- No active intervention is received
- They are given placebo to adjust for the placebo effect
- A benefit from a non-active substance

- The two groups are compared, at the end of the study, based on the variable(s) studied

- Experimental group

- Participants are chosen at random to be part of..
**Null hypothesis**- In a study
- There is no significant difference between the groups studied
- For instance, that a group receiving an experimental antidepressant, and a group receiving placebo, did not differ significantly on depression scores

- There is no significant difference between the groups studied

- In a study
**Alternative hypothesis**- In a study
- The difference between groups is due to difference in a variable under study, not randomly

- In a study
**P-value**- A statistical measure
- The probability of a study’s results, assuming that the null hypothesis is true
- In other words, how likely the relationship between the studied variables is due to simple chance

- The lower the p-value, the greater the chance that the alternative hypothesis is true
- The null hypothesis would then be false

- For instance, a study on the efficacy of an experimental antidepressant produced p=.02
- If the antidepressant had no effect, the difference observed between groups would come up in 2% of studies, due to an error in sampling
- So only 2% of studies would affirm the null hypothesis
- That there is no significant difference between groups

**Observational studies**- The researchers do not add an intervention
**Longitudinal**- The same participants are followed and evaluated over a significant period of time
- Better than cross-sectional studies at affirming causation
- Able to study the developmental trends of individuals
- Expensive
- Time-consuming
- Participant dropout
- Over time, people decide to no longer participate

- Practicing effects
- Getting good at taking tests due to the test format, not improving what the test is designed to measure

**Cross sectional**- Compares different age groups at the same time
- Can evaluate many different variables at once
- Takes a lot less time than longitudinal studies
- Less expensive than longitudinal studies
- No participant dropout
- Can’t determine causation
- Cohort effect
- A group of people who’ve been collectively influenced by certain societal or cultural events or circumstances

- Individual trends of development?

**Sequential design**- Both longitudinal and cross-sectional
- Several age groups studied over a period of time

**Naturalistic**- Within the natural environment of the participants

**Types of validity****External**- How much can findings in a research study, be generalized?
- Population
- How well does the sample studied, represent the whole population of interest in the study?
- How realistic is the process of finding participants of a desired type?

- Ecological
- How well does the experiment simulate the natural world that the participants live in?

**Internal**- Does the study do a good job at figuring out that the effect that’s being studied, is caused by what they think it is?

**Test**- Criterion
- Determining the relationship of a current score, and a future outcome
- How measurement in a study influences the conclusions of a study
- Concurrent
- When test scores and measurements ocurr in quick succession, or at the same time

- Predictive
- How well a test score is predicted to influence participants in the future

- Content
- How matched a test is to what is being measured in the studya given test is to measure what the study is hoping to measure

- Construct
- How well a test measures what it is supposed to measure
- Convergent
- How well two different tests are similar enough in ranking participants

- Discriminate
- Showing that measures with no relation, have in fact no relation

- Criterion
**Face**- A weak measure of validity
- How much a study covers the concept it studies
- How relevant the study is
- How “good”, in a subjective sense, the study is

- A comprehensive guide of 13 biases, based on the principles of the Cochrane Handbook of Systematic Reviews of Interventions
- Found here.

Sources: social/societal mores/values/country of studypurityr-squared valueSources: Ben Komor, https://explorable.com, Dr. Laura Wray-Lake, https://research.collegeboard.org, http://www.statsdirect.com, http://blog.minitab.com, http://stattrek.com, https://www.iwh.on.ca, https://himmelfarb.gwu.edu, http://study.com/academy