You cannot reverse-generalize from a study to yourself.
The process of science and statistics goes in the opposite direction: collect a sample, then generalize to a larger population.

If I collect data about a bunch of people, then tell you the average tendencies of those people, I have told you figuratively nothing about you. I say "figuratively nothing" because it isn't technically literally nothing, but it is damn-near nothing.

What I have told you is a summary statistic of a sample.
We can use statistics to generalize that summary to a wider population and the methods we use result in some estimate of the population average with some estimate of uncertainty around that average (or, if Bayesian, some estimate and a range of credibility).

Height as an example

To see a simple example of this, imagine measuring height.
If we measure the height of thousands of people we can calculate the average height of the sample, which generalizes to give us a very confident estimate of the average height of the population. That precise estimate of the average height of the population tells you figuratively nothing about my specific height or your specific height. Unless we measure my height, we don't know it, and same goes for you.

All else being equal, our best guess about either of us might be that we are both "average height", but that guess will be incorrect more often than it will be correct. All these measurements result in a best guess that is still likely to be several off by several centimetres. I call that learning figuratively nothing: our ability to guess height is no better than it would have been if we were just eye-balling it.

I grant that we would "learn" that we should guess that I am taller than 1 centimetre and that I'm shorter than 1 kilometre. This is a level of accuracy that I call knowing figuratively nothing about my specific height. Granted, if we didn't know "human height" as a range, we could estimate it by measuring the sample and generalizing from the population. In that case, we would learn a general distribution of heights that humans could have. We still wouldn't be able to precisely predict the upper and lower limits unless the tallest and shortest humans were in the sample.

Human beings are not fungible

We cannot be swapped out and replaced by one another because individual differences exist.

Atoms are fungible: one atom of gold is equivalent to any other atom of gold. The same is true of molecules. Scientists can study fungible atoms and molecules and discover properties that generalize across them. We can derive laws, like the Ideal Gas Law, and learn something about the material under study.

Once we reach the cell, the subject under study is no longer perfectly fungible.
As we reach the level of the human organism, individual differences make the application of the general to the specific impossible.

When it comes to humans, the individual unit —a person— cannot be swapped out with the expectation that results will be identical. Psychological research, then, can tell us about "the human population", but not the individual human being. We can estimate a ballpark in which the vast majority of people are expected to score, but we cannot accurately predict the score of any specific individual person.

We don't learn about you or me or anyone in particular.
We learn about "people".

A psychology example

Study et al. shows that a state-or-trait X predicts better performance on task Y.

The sort of finding abstractly described above usually amounts to saying that someone ran a study where some people came in to their lab, these people did a questionnaire, these people did a task, and the authors ran statistics that found a correlation between scores on the questionnaire and scores on the task.

What does this tell us?

Finding a correlation at the group-level does not mean that the data for every participant followed this specific pattern. Indeed, it is likely that some people in the sample had the opposite pattern or no pattern at all! Just because there is a correlation in the whole sample doesn't mean that there is a correlation for each individual.

Keep this in mind when reading studies.
Published papers are (typically) about the average across individuals and you cannot infer anything about one person from what you learn about the average. If you could apply the average to the individual, everyone would be the same height!

You are different from other people in ways that are not measured.
There is no reason to think a particular average-based result applies to you personally. Psychology doesn't use methods that let us make reversed-inferences.

Can we live by statistics?

Honestly, you can't. You just can't.

The best you can do is look at a study, then consider whether it might apply to you or not, but then you have to live your life to sort it out. There is no guide-book to living. If you want that, you're better off looking into philosophy, particularly continental philosophy, which is more about how to live.

If this still seems counter-intuitive, consider height as an example again.
Imagine asking, "Now that I understand that the average height may not apply to me, how can I use height research to find pants that fit me?"
You can't. To discover whether clothing fits, you just have to try it on for yourself.

The Experiment of Life

Even though you cannot apply averages directly to yourself, if you want, you could take studies as starting points for ideas to try.

For example, start off by measuring your mood for yesterday when you wake up. You do this for a few months to get a baseline. This gives you a baseline sense of how your mood generally is and also how variable your mood is.

Then start introducing little life experiments.

Maybe you read something about gratitude and you decide that's worth trying. You introduce a 5 minute gratitude practice into your day and keep recording your mood. After a month or two of doing that, you compare your mood to your baseline. If the gratitude practice seems to help, you can keep doing it. Otherwise, you may decide to drop it and experiment with something new.

Sure, there could be all sorts of confounds.
Maybe it went from winter to spring and you like spring better. Maybe it went from fall to winter and you dislike winter. Note down what you notice and maybe re-try in a new season if you think the seasons matters.

That is the method, though: measure, try, measure again, try something else, repeat.

After a couple years of daily measurements, you could notice longer-term mood shifts simply by plotting the data as a figure. You could see seasonal shifts once you've got a couple years of data. With enough data, you could take seasons into account when interpreting new self-experiments.

The experiment of life is personal.
Interventions only apply if they work for you. There is no way to tell, by trying an intervention on other people, whether or not that intervention will work on you. Human beings are not fungible. You just have to try for yourself.

Experimenting with Clinical Findings

If you are looking for applications, look into clinical psychology.
For example, there are plenty of things you can apply in your life in Acceptance & Commitment Therapy (ACT). You can find workbooks online and you can try using those to improve your life. Research showing beneficial effects in samples doesn't mean these interventions will work for you, but they provide evidence that they worked for some people, which you could take to mean they're worth trying.

Conclusions

We might have to lower our expectations of what we learn in psychology.
Unfortunately, many ideal forms of information probably won't be discovered or created in our lifetimes.

This doesn't mean we can't have a good time!
We can live great lives, but we'll have to sort out what works for us for ourselves.
We cannot find the answers in scientific articles.
Articles can provide starting points, but not answers.

Index

Return to Statistics

Jump to Effect-Sizes