Full transcript
0:07Welcome to the fourth video in SPSS for beginners from the RStats Institute at Missouri State University.
0:15Now that we have learned how to examine each variable using descriptive statistics and graphs,
0:21I'm going to show you how to do some simple analyses. So this video will show you the basics of doing correlation.
0:28When you're ready to do a correlation for real,
0:31watch the other are stats Institute videos to learn more about the theory the analysis and
0:38how to write up your findings in APA style.
0:49We're still using the same
0:50SPSS data set that we created in the first video, however, I deleted the z-score variables that we created last time.
0:58Now I'm going to show you how to calculate
1:01correlations in
1:03SPSS
1:05The correlation that we are doing is called Pearson's r.
1:10A Pearson's correlation describes the relationship between two variables.
1:16Pearson's r ranges between -1 and +1.
1:210 indicates no relationship at all.
1:24The closer that the correlation is to either +1 or -1, the stronger the relationship
1:31between the variables.
1:34We are interested in the relationship between height and weight.
1:39Notice how the data have already been set up. Each person has a pair of scores.
1:45Your height should be paired with your weight, it makes no sense to pair your height with my weight.
1:52So it's very important that each pair stays together
1:56We have 10 pairs of scores. Our sample size is 10. Each pair counts as one case.
2:04So remember that we have two people without height and weight scores.
2:09They are not going to be included in this analysis.
2:13In fact, SPSS will simply ignore those cases with missing values. So let's do a correlation.
2:20Go to Analyze
2:23Correlate
2:24A Pearson's r correlates two variables, so choose Bivariate...
2:31As before all of our variables are here on the left. The two that we want to correlate are height and weight.
2:37So we need at least two
2:39variables.When you move over the first, the "OK" is still not available until you move over the second.
2:46And we could add additional variables, but each would be correlated only two at a time.
2:53We have some additional options here as well. We could calculate Kendall's tau or
2:59Spearman's Rho if we had different data, but for now let's just stick with Pearson's r.
3:06SPSS assumes that we want two tailed significance tests and
3:10that we want to flag significant correlations.
3:14We haven't talked about significance tests yet,
3:16so for now just know that significance tests tell us something important about the variables. In this case, our
3:24correlation is statistically significantly different than 0. If it is,
3:29SPSS will flag it. All of the default settings are just the way we want them, so click OK to run the analysis.
3:37The box that we see is called a "correlation matrix."
3:41The correlation matrix shows the correlation coefficient for every combination of variables.
3:48So we have two rows: one for height, one for weight.
3:52And we have two columns: one for height, one for weight.
3:56Where each row and column intersect, we see the correlation coefficient between those two variables.
4:05So in this quadrant of the matrix we see the correlation coefficient between height and
4:11itself.
4:13No surprise. It's 1. It's a perfect correlation.
4:17We see another perfect correlation down here on the lower right which is the correlation between weight and itself.
4:25SPSS will compare every combination of variables
4:29including each variable and itself.
4:32Now these correlations are not very interesting because we already know that
4:38every variable will always correlate with itself at a +1, no matter the variable.
4:45The interesting correlations are in these off diagonals.
4:50The top left box is the correlation coefficient.
4:53It will always be between +1 and -1.
4:58Below that is the significance level.
5:01Significance levels smaller than .05 are statistically significant.
5:06Below that is the N, or the sample size, which is our 10 pairs of scores. So let's look at this coefficient.
5:16Notice that the off diagonal correlations are the same because height correlates with weight exactly the same as weight
5:22correlates with height. In this case it's a .574 which is pretty strong,
5:29but not significant because the sample size of 10 is pretty small.
5:35You are always more likely to find significance with larger sample sizes.
5:40If this correlation was significant, we would see some asterisks next to the coefficient.
5:47So as I mentioned, you can correlate more than two variables at a time,
5:51and you could even use correlation with nominal variables as long as it only has two levels. In fact, let me show you.
5:59Go to Analyze
6:01Correlate
6:03Bivariate
6:05All we're going to do is throw in a third variable, Gender, and this is actually called a point biserial correlation,
6:14more on all of that later. For now, just click OK.
6:19We get another correlation matrix, but this time it's bigger. It has three rows and three columns.
6:26The correlations between height and weight are exactly the same as before,
6:31but we also have correlations with gender.
6:35Because the correlations are negative, as one variable goes up the other goes down.
6:41Remember that we coded males as 1, females as 2. So the 1 is smaller. We see this negative
6:48correlation, the smaller numbers are associated with larger values. So basically the males were taller and weighed more.
6:58And here we also see a significant correlation that's been flagged.
7:03The biserial correlation between weight and gender has two asterisks, so what does that mean?
7:11We can see that this correlation is significant at the .01 level. In fact, the p-value is .009.
7:21So there' is a statistically significant relationship between weight and gender.
7:26So there's one more thing that I want to show you with correlations,
7:29and that is how to make a picture of them. The picture is called a scatter plot,
7:35and it is created using a new tool called the chart builder. And here's how we do it.
7:42Instead of the Analyze menu, we're going to use the Graphs menu. So go to Graphs
7:49Chart Builder
7:51We will learn more about the chart builder later when we learn about graphing.
7:56For now, let's just have some fun and make a scatter plot.
8:01Start by clicking on the word Scatter/Dot in the gallery.
8:07Now we see our eight options. If you hover your cursor above them, SPSS will tell you what they are.
8:16We want this first option: Simple Scatter
8:22Click and drag it into the blank area known as the canvas.
8:27You will see that we now have two drop zones: one for the x axis and one for the y axis.
8:33So let's use height to predict weight.
8:37Drag height to the x axis drop zone
8:42weight to the y-axis drop zone.
8:46And that is all you have to do. Click OK.
8:51And there is our scatter plot of all 10 of the pairs of scores.
8:57There is much more that we could do with correlation, so for instance, we could format the scatter plot in APA style.
9:06We could do other types of correlations.
9:09We could even use some variables to predict other variables using a technique called
9:15regression
9:16To learn more about correlation, scatter plots, simple regression, and multiple regression check out these other videos
9:24from RStats Instiutue.
9:27Correlations are about relationships between variables,
9:31but we might also be interested in
9:35differences between variables. So next we're going to learn about t-tests.