If you're seeing this message, it means we're having trouble loading external resources on our website.

If you're behind a web filter, please make sure that the domains *.kastatic.org and *.kasandbox.org are unblocked.

### Course: Statistics and probability>Unit 6

Lesson 4: Types of studies (experimental vs. observational)

# Types of statistical studies

Types of statistical studies.

## Want to join the conversation?

• At Sal puts a restriction on the control group of the experiment. Doesn't that take them out of the control group? It seems like in this example we now have two experiments, but no control group. From my understanding of control groups, they either just keep on doing what they're doing, or the changes they make have a known effect, e.g., current best medical treatment (control) versus experimental treatment.
• Here's what I got from this video, and I hope this helps: what is done with the control group and the experimental group will vary on what you're testing for.

With this experiment example, Sal is trying to see if the amount of computer time has an impact on blood pressure. You can think of the reason for the experiment is that, in the observational study, we saw a positive correlation between computer usage and blood pressure. So the question for the experiment is: "does high computer usage time cause high blood pressure (is there causality)?"

Since we want to see if there is causality between high computer usage and high blood pressure, we need then one group, the control, to have a low computer usage, and the experimental group, what we're testing for, to have higher computer time usage. We do this test this way because we need to control the amount of computer time each group has, or we won't have an experiment.

If both groups, the control and experimental groups, were not given different restrictions on computer time, the test would only be showing what the blood pressure was for each group afterwards, and we wouldn't be able to tell if computer time made any difference.
• So what EXACTLY is a confounding variable?
• A variable that's not accounted for that may cause variation in the results. You might a conduct a study and conclude that bank thieves are more likely to eat ice cream after the theft. While your results may be statistically robust, you have overlooked the fact that it's because most thefts happen in the summer when the weather is warmer. In this case, "season" or "temperature"are the confounding variables.
• I'm confused about the difference between the sample and observational study examples. Suppose in the sample study, they collect computer time and blood pressure. It sounds like if they just present the average of both then it remains a sample study, but as soon as they plot them against each other it becomes an observational study. Similarly, in the observational study, they presumably created some criteria to determine which 1000 people were included in the study. Is the difference really based on what is done with the data rather than how the study is conducted?
• The type of study you use really depends on the data that you have, and what you're trying to find out.

So, your first question should be is, what am I trying to find out? Am I trying to find out something about a population? Am I trying to compare two variables (like computer use and blood pressure)? Am I trying to see if something causes something else?

Then, the second question to ask is: what kind of data do I need to collect so that I can answer what I am trying to find out?

I hope also that you continue with the next few videos, where Sal works through examples of each different type of study. That will help I believe in telling the difference between them.
• Ethical experiment: Trying to determine if computer time causes high blood pressure
Unethical experiment: Trying to determine if high blood pressure causes computer time
• Basically, you can't ever really know you've eliminated the confounding variables for lots of different scenarios.

How would you know? Looks like it depends on the scenario being studied.
(1 vote)
• Eliminating confounding variables entirely can be challenging, and it depends on the specific scenario being studied. Researchers use various techniques such as random assignment in experiments to minimize the influence of confounding variables. However, in observational studies, confounding variables may still exist, making it difficult to establish causal relationships definitively.
(1 vote)
• What would be the term for a statistical study in which only one variable (for instance, it can be heights of male giraffes in a certain forest) is studied to find population parameters, and the entire population is studied, instead of just a sample?
• Can someone please make a summary of this whole video?
• Sal explains three main types of statistical studies: sample study, observational study, and experiment. In a sample study, researchers aim to estimate a population parameter by sampling a subset of the population. An observational study explores the relationship between two variables within a population, without establishing causality. Confounding variables may influence the observed correlation. Experiments, on the other hand, aim to establish causality by manipulating an independent variable and observing its effect on a dependent variable. Random assignment helps minimize the influence of confounding variables. Sal emphasizes the importance of distinguishing between correlation and causation, and the need for careful experimental design to draw meaningful conclusions.
(1 vote)
• I understand about what the type of studies are but something is not working in Sal's experiment.

Sal said that he will see if computer time will affect blood pressure. He said experiments are to see if there is a cause, but isn't there a confounding variable there? The people on 2 hours on their computers are less active so couldn't the difference between the 2 groups' blood pressure be because of less active time?