If I am taking measure of of a subject coming into a program, whereby exposure to the environment (over the course of years) will be the treatment, with post testing using baseline metric on intake, would it be prudent to classify main themes within the treatment as variables, and if so, should these variables be directly related to the metrics in previous studies?
For example, measuring personality coming into a program, identifying potential personality influences with in that environment, and testing on coming out of the program.
Or do I take both pre and post measure and then go back to the environment and speculate on which variables likely played a role?
Dear Austin,
Pre-post design could be good for a small n size. Then you use non-parametrics statistis or like we did
https://en.wikipedia.org/wiki/Nonparametric_statistics
Article Depression, anxiety, hostility and hysterectomy
Considerations for Tracking Environmental Influences on Variables in Small N Data Design
A. Questions Concerning Initial Data Research.
1. Given the initial variables tied to the hypothesis and null-hypothesis, are you considering counterfactual explanations?
2. How are you defining early warning indicators that cue an unexplained shift in dependent variable behavior? Are indicators rolled in as alternate variables (intervening, confounding, endogenous, exogenous, moderating, etc.)?
B. Consideration No. 1: Multiple Hypotheses Analysis
1. Assumptions. Measurement types vary with variable properties: measures of association and measuring patterns of relationships (direction, strength, significance, etc.) will differ based on mixing and matching different variable types (ordinal, interval, ratio?)
2. Suggestion – let the data speak for itself. Could we establish a cross-sectional analysis using a simple cross-tabulation table where counterfactual explanations or unforeseen influences are eliminated (or not!) by looking at measures of association and significance?
C. Consideration No. 2: Bayesian Analysis
1. Assumption. Bayesian analysis could address the question of what is the probability of a causal relationship given a change in conditions.
Can Bayes formula be applied as indicators suggest a behavior modification that appears to covary with a certain set of environmental influences? Could you apply this formula to something simple like MICROSOFT EXCEL and enter your observations of variables and calculate daily the probabilistic outcome to identify possible shifts in patterns?
Forumla: P(Hi|D) = P(D|Hi) x P(Hi)/P(D) where [P= “probability of”]; [| = “given”]; [/ = divided by”]
Problem-Set: measuring personality coming into a program, identifying potential personality influences with in that environment, and testing on coming out of the program.
P(Hi): possible outcome (hypothesis) – likelihood of a behavior given a certain environmental influence or set of environmental influences
P(D): probability of environmental influence(s) occurring
P(D|Hi): the probability of environmental influences occurring given that the hypothesis
P(Hi|D): calculating the posterior odds – what is the probability of behavior modification given the environmental influence.
- Badges
- Science topic