Tip:
Highlight text to annotate it
X
Hello, my friends.
We near our--
the end of our journey through ANOVA.
In this video, we're simply going to look at the protocol
that one might follow as one conducts an analysis of
variance and reports on it in their research.
Remember that the ANOVA is about two or more groups
looking at one variable.
In other words, group one has this distribution of the
variable, group two this distribution, group three that
distribution.
I'll remind you that we have normality.
Many researchers, however, point out that normality is
not required.
The study is still robust.
They will say that you need to have approximate normality.
I think that's the way of saying it.
Bear with me.
And we're going to go through the protocol and then look at
the SPSS output readout to discuss each of the issues
that we would find there.
The protocol for conducting an ANOVA is as follows.
And the researcher should examine each of these.
Now friends, this is my suggestion.
There is no one right way to do this.
There are, however, wrong ways.
I hope you will consider this one of the right ways.
The first thing that you need to look at is normality in
your data set, in your dependent variable.
Kurtosis and skewness are the means by
which you examine normality.
Then you need to look at the homogeneity of variance.
That means do the individual curves have the
same standard deviation?
That would be done with a Levene's test.
You would identify your descriptives.
You would conduct your ANOVA.
You would do post hoc as needed.
If you find significance in your ANOVA, then you would
conduct your post hoc.
And if you found significance, I think that it would be
important that you evaluate the effect size and then that
you also examine the power of the study.
Well, hello friends.
Now we're going to move on and actually look at the SPSS
readout, where we collect the things that are necessary for
us to meet the ANOVA protocol.
Now you recall that the first thing we wanted to do was a
normality, where we looked at kurtosis and skewness.
If you'll recall, to do kurtosis and skewness you go
to frenquencies.
You should have something comes up, look like this.
The skewness is a negative 0.842, a
little bit to the right.
Kurtosis, 1.924.
This isn't far off of a normal distribution.
Let me--
It's not normal, but you know, it just simply--
it needs to be approximately normal.
And the study is still robust.
Then we wanted to conduct the one-way ANOVA.
We collected our descriptives.
We look at homogeneity of variance.
The Levene statistics indicates to us that we do not
have homogeneity of variance.
It's not there.
That's an assumption that's not met.
And we would need to tell that to the informed reader.
When we conducted the ANOVA between groups, we note that
there is a significant difference.
I want to point out to you that the
significance is not 0.000.
The significance is less than 0.001.
The probability that this--
that it be zero is--
that's just not possible.
And of course, the computer puts it at 0.00.
But actually, you would report it as less than 0.001.
So it is significant.
Once we do that, we would need to have the post hoc analysis.
We notice that group one, group two differ.
Group one and three do not differ.
So the private differs--
the public differs from the for-profit.
And the public does not differ from the private
not-for-profit.
The for-profit differs from the public.
And then we notice that the private not-for-profit doesn't
differ from either one of them.
So you've got some things that are there.
Once we have done that, and we've got it, we wanted to go
in and see if we could find the effect size and the power.
If you'll recall, then we went out to the comparison, looked
at this, got us a univariate analysis of variance.
And let's see if we can get down there and find the
partial eta squared.
Here we are.
The partial eta squared was 0.67, which means we have a
very strong effect size.
Our power is 0.995.
That is an awesome, awesome power.
So in the readout that you have here, I went through and
I did several tests.
I did the test for normality using the frequencies, the
homogeneity of variance, the descriptives, the ANOVA.
The post hoc was all done in the ANOVA.
And then when we did the effect size and power, we went
in and did the general linear model and obtained those
statistics.
So now I've got a good readout that has everything that I
need in order to answer the required protocols.
Well, I think we have made it through ANOVA.
This is the last of the video.
The last one was to look at the protocol.
I raise my cup to you.
You've done well, pilgrim, to come so far with so much hair,
with so many after it.
If you will follow that protocol, you will have a
professional-looking study if you do ANOVA.
May the odds be ever in your favor.