This is a five-section course as part of a two-course sequence in Research Methods in Psychology. This course deals with experimental methods whereas the other course dealt with descriptive methods.

Loading...

Del curso dictado por Georgia Institute of Technology

Experimental Research Methods in Psychology

calificaciones

This is a five-section course as part of a two-course sequence in Research Methods in Psychology. This course deals with experimental methods whereas the other course dealt with descriptive methods.

De la lección

Introduction

- Dr. Anderson D. SmithRegents’ Professor Emeritus

School of Psychology

Hi, Anderson Smith again.

In most of the experiments we talked about,

we're dealing with one independent variable and one dependent variable.

There are many situations, however,

where we want to use more than just one independent and one dependent variable.

And now designs can do that.

In fact, they can be very rich designs giving us a lot of information about

what's going on.

For example, many times you want to have multiple dependent variables,

where we have different measures of the same construct.

For example, if we're going to look at stress, we could look at arousal,

how aroused we are, that's a measure of stress, cortisol level,

which you can measure very simply, just self-reports, how stressed are you?

We can look at those and that helps us, for example, one measure of

these things that we all know measure stress can be used as a manipulation and

then another one can be used to simply check with the manipulation.

Is the manipulation that we are making really doing what we say it's doing, and

that's called manipulation check.

And we can do that if we have multiple measures of the same construct,

multiple measures of the same dependent variable.

We also might want to just look at different

constructs measured in the same experiment.

We might want to look at stress, but also anger and worry and anxiety.

Other things which have different definitions than stress might covary in

the same way.

So in many situations where you might want to use multiple dependent variables.

Likewise in many situations where you might want to use multiple independent

variables, manipulate more than just one thing in the experiment.

And we can do that.

In fact, it's more efficient to measure multiple independent variables in the same

experiment.

If we have four independent variables we're interested, it's more efficient than

in one experiment, they have four different experiments.

And those are called factorial designs.

Designs that have factorial manipulations of many independent variables

at the same time.

And the advantage of multi-variable experiments is if you can look at the main

effects of each independent variable like we would do an experiment with

just one independent variable, but now with multiple independent variables,

we can also look at interactions among them.

That is, the effects of one independent variable might really depend

on one of the other independent variables.

And we look at those interactions between variables when we have a multivariable

experiment.

Now each level of one independent variable is paired with each level

of another independent variable, and that creates a condition of the experiment.

The way you do that, let's say we have three variables, one that has two levels,

another that has three levels, and a third that has two levels.

We want to know how many conditions we have, we simply multiply those together.

So two variables, two levels of one variable, times three levels of the second

variable, gives us 6, times two levels of the third variable, gives us 12, so that

12 different conditions in the experiment that we would assign subjects to.

One representing each level of each variable.

And when we do that, then we can look at interactions

as well as main effects in our statistical designs.

Now, we do have to randomly assign participants to each condition.

We do that with all experimental designs.

But we know now how many conditions we have to have.

With three variables, with one with two levels, one with three levels,

one with two levels, we would have 12 different conditions.

Let's look at an example.

Here's a factorial design with two independent variables,

each having two levels.

So that means two times two is four different conditions.

We have two levels of the first independent variable, A and B, and

then two levels of the second independent variable little a and b.

So we have four conditions.

And we would randomly assign

participants to each of these four conditions in order to do the experiment.

And now we can look at the main effect of independent variable 1 in a statistical

analysis, the main effect of independent variable 2, and the interaction.

Does the effect of independent variable 1 or

independent variable 2 depend on the effect of the other independent variable,

that interaction sometimes gives us the most important findings in an experiment.

Let's give an example.

Let's say we want to study the effects of room temperature, independent variable 1,

and light intensity, independent variable 2, on student test taking.

Their performance on tests.

Very applied question, but it can be asked.

So we have two levels of room temperature.

Let's say 72 degrees and 92 degrees.

And then we have two levels of room illumination,

let's say the regular lighting in the classroom, and

then very bright lighting by bringing in a bunch of floodlights.

So we have four conditions.

And now we do the experiment and we find that both are important.

Room temperature produces a big effect, and even lighting produces effect.

But they have sort of different kinds of effects.

And that different kinds of effects is reflected in the fact that there's

a significant interaction.

That the effects of lighting depend upon the effect of room temperature.

Let me just plot that out to show you what I mean.

Remember this is a hypothetical study that we're doing here.

So we have regular or bright lighting, lighting level,

we're measuring test performance and we have two temperatures.

And what we found that if temperature 72, the effective of lighting is very small.

If the temperature is 92, the effective lighting is much more dramatic.

So we have bright lights when the room is hot.

That produces a greater decline in performance.

And they both are significant, both in they decline with the lighting and

decline with the temperature, but the interaction is the big effect here.

That the decline with lighting level really depends upon the temperature

of the room, the interaction.

Interactions can be absolutely the most important way to look at what the effects

are in the experiment.

Thank you.

Coursera brinda acceso universal a la mejor educación del mundo, al asociarse con las mejores universidades y organizaciones, para ofrecer cursos en línea.