Planned Contrasts SPSS example PDF

Preview

Summary

planned contrast test notes...

Description

Planned Comparisons: SPSS Procedures and Results This file will cover how to run Post Hoc tests through the Analysis of Variance (ANOVA) function in SPSS. The file we will be using is titled “hourlywagedata.sav” which includes data from nurses. In this file, there are four variables; we will be focusing on two: “agerange” and “hourwage”. We will be running an F test with post hoc tests to determine if different age groups earned different amounts of money. To run planned comparisons or contrasts, run a one-way ANOVA, follow the steps presented in Unit 2. In addition, when “Univariate” window is open, and after you’ve entered the independent and dependent variables in the appropriate fields, select “Contrasts…”. You should see the following window:

What you will see is that no contrasts are selected yet, as it says “None” in parentheses after the “agerange” variable. To select a type of constrast, you need to choose from the drop-down menu where “Constrast:” is labeled. There are several options available for contrasts, and they are described in the PowerPoint slides. Here, we will select “Simple” contrasts, which will compare the means of both group 1 and group 2 to the mean of group 3 (but will not compare the means of group 1 and group 2 to each other). You can select whatever type of contrast you like, but once you select the type of contrast, you must select “Change”. When you do so, “None” will change to “Simple” or whatever contrast you chose:

Next, select “Continue” and then “Ok”. You should then see the output file. The output should look the same as the output for Unit 2. However, there will be a couple of additional tables at the end:

Custom Hypothesis Tests Contrast Results (K Matrix)

Age Range Simple Contrasta Level 1 vs. Level 3

Contrast Estimate Hypothesized Value Difference (Estimate - Hypothesized)

Dependent Variable Hourly Salary -1.643 0 -1.643

Std. Error Sig.

Level 2 vs. Level 3

95% Confidence Interval for Lower Bound Difference Upper Bound Contrast Estimate Hypothesized Value

.394 .000 -2.417 -.869 -.504 0

Difference (Estimate - Hypothesized) Std. Error Sig. 95% Confidence Interval for Lower Bound Difference Upper Bound

-.504 .283 .075 -1.059 .051

a. Reference category = 3

Because we are comparing 3 different age ranges groups (18-30, 31-45, and 46-65), there are two contrasts provided (K – 1). The first contrast compares the means of group 1 and group 3. The value of the difference between the means is -1.643, and the significance level is .000, indicating that there is a statistically significant difference between the means of the two groups. The second contrast compares the means of groups 2 and 3. The value of the difference is -.504, which is not statistically significant ( p = .075). As you can see, this planned contrast does not compare groups 1 and 2. This is one of the downsides to planned contrasts; you can only conduct 1 fewer than the number of groups you are comparing. Thus, you want to choose the appropriate

contrast wisely. However, the advantage is that planned comparisons have greater statistical power than post-hoc tests. Of the contrasts available in SPSS, most are fairly similar, but simply change which groups are being compared (and some compare multiple groups simultaneously). The one that is a bit different is the “polynomial” contrast. This contrast can be used to examine the shape of trends. This is only really useful if the ordering of the groups, or levels of the independent variable, matter. Fortunately, with this data set, they do, since they are based on age groups. If you run polynomial contrasts, you will obtain the following table: Contrast Results (K Matrix) Dependent Variable Age Range Polynomial Contrasta Linear

Contrast Estimate Hypothesized Value Difference (Estimate - Hypothesized)

Quadratic

Hourly Salary 1.162 0 1.162

Std. Error Sig. 95% Confidence Interval for Lower Bound Difference Upper Bound Contrast Estimate Hypothesized Value

.279 .000 .615 1.709 -.259 0

Difference (Estimate - Hypothesized)

-.259

Std. Error Sig.

.209 .216

95% Confidence Interval for Lower Bound Difference Upper Bound

-.670 .152

a. Metric = 1.000, 2.000, 3.000

Because we are comparing 3 groups, we can only conduct linear and quadratic contrasts. If we had a 4th group, we would have a cubic contrast as well. Basically, these contrasts can be used to determine if the trend is a straight line, or some sort of curved line.

When I say “straight line,” I’m referring to the hypothetical line that connects the means of each group:

You can obtain a graph of the means by selecting “Plots…” in the “Univariate” window, and then placing “agerange” in the horizontal axis (make sure you click on “Add”). As you can see from the plot, the trend is almost a straight line, with a slight curvature. According to the results of the contrasts we ran, a linear trend significant fits the data. Our contrast value was 1.162, p = .000. While there is a slight curve in the line, the test of a quadratic relationship was not statistically significant, with our contrast value of -.259, p = .216. So, we can assume the trend is linear in the population....