Monday, 15 April 2024
------------------------

Hello. All is well. Continuing with chapter two, we look into correlation, experimentation, and ethics.

Correlation describes the relationship between two variables. Correlation coefficient (-1 to 1) tells us the direction and strength of the correlation. A negative number means negative correlation, that the variables move in opposite direction, while a positive number means positive correlation, that the variables move in the same direction. Zero means there's no correlation. The coefficient can be any number between -1 and 1, defining the strength. It's like valence (direction) and arousal (strength) for affect. We can display the correlation between two variables visually with a scatterplot, plotting each entry of two variables (x, y) we have.

But, correlation does not mean causation, that there's a cause and effect between the correlating variables. It could be a third variable that both variables depend on, known as a confounding variable. An example is increase in temperature causes an increase in ice cream sales and drowning deaths. We may think eating ice cream cause drowning, but in actuality, it’s due to more people being in waters because of the hot temperature. We're also prone to confirmation bias, our perception is learned and our brain seeks patterns, making us see illusory correlations.

In order to confirm causality, we first make a hypothesis of cause and effect, specifically an experimental hypothesis, that we can then design an experiment for. We have an interdependent and a dependent variable. In our experiment, we want to see if a change in the interdependent variable alone will have an effect on the dependent variable. To do this, in its most basic form, we set up two identical environments, an experimental group and a control group. Then, we manipulate the interdependent variable in only one of the groups, the experimental group, and check if there's a difference between the dependent variable of the two groups. It's important that our experiment is understandable and repeatable. We do this by providing an operational definition of the experiment, a description of the variables.

To avoid placebo effect or expectation bias for participants, we ensure they don't know which group they're in, known as single-blind study. But, we can also avoid expectation bias for researchers, those in control of the experiment, by making them unaware which group participants are in too, known as double-blind study. I imagine double-blind study is useful when the measurement of the dependent variable is more subjective or qualitative.

Now making two identical groups to isolate the interdependent variable is easier said than done. It's preferred to have a random sample, a subset of a larger population, and also a random assignment, randomly assigning the participants' group. The goal is to ensure that there's no significant difference between the two groups other than the interdependent variable and that the sample can be generalized to the larger population.

We call a study quasi experimental if the interdependent variable is not fully isolated (it can't be manipulated alone), for example, a person's sex is not fully isolatable, you can't change a person's sex. That's at least what I understood. It does make sense though, biological sex is usually accompanied with a specific social role (gender), therefore the person's social role may be the cause, not the person's biological sex.

Researchers share their findings in peer-reviewed journal articles. Before they can be published, other scientists will review the article to ensure it's all goody good. For quality control, the editors of the journal will collect all reviews to determine if the article can be published, needs revision, or should be rejected.  There's a bunch of checks of validity, such as ecological validity (findings are generalizable), construct validity (the measurement used actually measures what it's said to measure), face validity (face value of the experiment actually testing what it claims to test). I imagine that you must know which journals are credible, that have a good track record. Apparently there has been an issue lately with experiments not being reproducible. This is inline with the emotion book, where they talked about issues with replicating experiments that found innate emotional expressions. Anyhow, APA has a manual for how to write your article.

Experiments by institutions with federal support must have a board to review research proposals. Human participants should not be harmed, and must sign an informed consent form, detailing potential risks and implications. To preserve the integrity of a study, deception of what the study is about may be necessary, but after the conclusion, participants must be given a full debriefing of the study. Studies with animal participants, usually rodents and birds, have less scrutiny but pain and suffering must be minimized.