3.2.1 Observational Studies and Experiments

Added: 2026-06-03

[00:00:00]in this section of notes we're going to look at the difference between an observational study and an experiment the main difference here is how you go about collecting your data so in an experiment you have to do something you have to pick one or more treatments often including a placebo or a control and you have to impose that treatment on your subjects whereas an observational study you're just watching what happens and recording data about whatever it is that happens so there's no treatments in an observational study one thing that you should add to your notes here is that there are two kinds of observational studies an observational study can either be prospective or retrospective so in a prospective study that just means that the investigators design the study recruit the subjects and collect the baseline data before the subjects develop whatever outcome they're interested in so they're they're getting new data whereas in a retrospective study that means that the investigators are using data that already exists somewhere so the outcomes have already happened so the difference between these would be let's say that you wanted to investigate whether or not playing a musical instrument has an impact on grades if you just went out and talked to some seniors and said ask them questions like how long have you been playing a musical instrument do you play one at all and what's your current GPA that would be retrospective you're doing that after the fact they've already gotten their grades and they've already played an instrument or not played an instrument prospective would be maybe you would pick some kindergarteners and you would follow them throughout their high school career and you would pay attention to did they play a musical instrument how many years did they stick with it what kind of grades did they end up getting so it's a very small difference between prospective and retrospective it's just when you pick the people to be in your study and whether you're looking the data as it happens or if you're looking at data that already exists in happened in the past now we've talked about explanatory and response variables so in that example I just gave you a moment ago the explanatory variable would be whether or not the student played a musical instrument and then the response variable would be the student's grade point average now the problem here is that in most observational studies there are a lot of confounded variables so that would be something else that maybe relates to whether or not a student plays an instrument and also relates to their grade that kind of makes your results misleading it makes it hard to tell whether or not it was the musical instrument that caused the increase in the grades or something else so for example maybe students that have parents that are highly involved in their schooling maybe they're more likely to stick with playing a musical instrument for an extended period of time and maybe they're also more likely to get better grades because their parents are checking their grade books online and that sort of thing so that would be an example of a confounding variable would be how much the parents are involved in their students lives so now let's move on to looking at experiments and the pieces of information that go along with that so an experiment is when we have our subjects and we have to give them some sort of a treatment so I can take the example that I gave you before and say this time I'm gonna pick all of these kindergarteners and I'm going to decide through random assignment that these ten kindergarteners they're gonna be required to play a musical instrument the whole way through school they're not going to be allowed to quit they have to keep with it and then there's other ten students they're gonna be forbidden from playing a musical instrument so that would be the treatment is whether or not they are assigned to play a musical instrument the experimental units would be those students also sometimes called the subjects now there are some things that can go wrong so let's look at another scenario here let's say I'm gonna take my experimental units I'm gonna give them a treatment I'm gonna see what happens in the end so let's take the example of whether or not a student takes an online SAT course and how that impacts their SAT score now there are some other things that might impact how an A student does on the SAT other than just whether or not they took a course maybe the students have different course histories so maybe you've got some students that have taken all the way up through calculus because they're seniors or maybe you've got other students who are just taking Algebra one so they're still at the beginning of their career that might be a confounding variable would be their history of math classes so to control for this what we want to do is we want to randomly assign students and if we randomly assign them to the groups then that confounding variable is going to end up being lessened so we'll end up with some older students and some younger students in each of the groups which means we'll also end up with some students that have had more or less math classes in each group so that random assignment is really important so that it makes the two groups as equally likely as possible