How to run the GLM Repeated Measures procedure

1. Build the Data Set

The Experiment (the translation priming study): We wanted to know if an L1 translation presented very brief and masked would facilitate the recognition of an L2 word. We presented English letter strings on a computer monitor and asked Chinese-English bilingual speakers to decide whether the letter string was an English word or not. A target word was preceded either by its L1 translation or an unrelated L1 word (the preceding word is called a prime word). We also tested native English speakers on the same materials. This is a mixed design with one within-subject variable (prime-target relation) and one between-subject variable (language background). Each variable has two levels, translation vs. unrelated for the former and native and non-native for the latter. Thus, we use MLM repeated measures to analyze the data.

The data set should look like this.

1. For subject analysis, each case represents a participant.

2. The between-subject (grouping) factor is in the second column; NS and NNS indicate two different groups. Use one column for the grouping factor regardless of the number of the level of this factor. It can be NS, NNS1 and NNS2 when nonnative speakers of different proficiency levels are included in the experiment.

3. Each measure/level of the within-subject factor takes
a column.

4. For item analysis, each case represents a test item; (Make sure you copy the item data, not the subject data, from the output file from the update command of DMASTR.)

2. Run the Procedure

Start SPSS, under Analyze, click General Linear Model, then Repeated Measures, you will see the following dialogue box (1):

.(1).((2) ......

1. In the first textbox, type in a transparent name for the within-subject variable, e.g., relation;
2. type in the number of levels/conditions of the within-subject factor, in this case, 2; (See (2))
3. Click on Add to paste the factor name and number of level to the lower box (see below);

4. Click on Define to open the following dialogue box. There are four variables shown (see the data set above).

5. Define the conditions by highlighting and moving within-subject variables (trans & unrel) into the Within-Subjects Variable space and the between-subject variable (lbackg) into the Between-Subjects Factor box (see below).

6. Click OK. (Click on the Options button and check Descriptive Statistics and then Continue if descriptive statistics is needed)

3. Reading the Output

1. First look at the within-subject effects table.
a. There is no main effect of the Relation variable (F=1.11; p=.306)
b. There is a significant interaction between the two variables Relation/Language Background (F=7.385; p=.014)

2. Now the Between-Subjects Effect.
a. There is a main effect of language background (F=48.521; p.000)

A Three-Variable Example

The above example has one within-subject variable (prime-target relation) and one between-subject variable (language background). If you have two within-subject variables, you have to be careful about the order of listing and entering the variables. Suppose we want to know if case is indeed an important factor in word processing for non-native speakers from a non-alphabetic language. We test three groups of people, native speakers of English and French and Chinese ESL speakers. The materials include the same words presented in upper case and lower case on two counterbalanced lists. We also include high and low frequency words in each case. So we have one between-subject variable (language background) with three levels (English, French, Chinese), and two within-subject variables (case and frequency), each with two levels (upper vs. lower cases; high vs. low frequency).

1. The Dataset

List the between-subject variable in the first column (only two levels are shown in the example: English and Romance). Use the next four columns to present the data from the two within-subject variables. Note the order of the variables: 1/1, 1/2, 2/1, 2/2, or uppercase/high frequency, uppercase/low frequency, lower case/high frequency, lower case/low frequency.

 

 

 

 

 

 

 

 

 

 

 

 

 

2. Analyze the data  
2a. Select the MLM Repeated Measures
procedure and you will see the dialogue
as shown above 2(1).
2b. Enter the "Case" variable first; Then "Freq". Both have 2 levels. Click Define.

 

2c. Enter the within-subject variable in the
same order as listed on the left by selecting
them and then clicking the right pointing arrow. Move the between-subject variable (lb) into the second space. (ignore the last four variables on the left, which are error rate data) Click OK to run the analysis.

You may want to click on Options, and check Descriptive Statistics.

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 


3. Reading the output

3a. Descriptive statistics
You can use the means in this table when you
present the results in a table in your report. You
need to reformat the table so that it is more
reader-friendly. Check a few published articles
to see how the results are presented in tables.

3b. Within-subject main effects and interactions
The main effect of each within-subject variable and interactions can be found in the Tests of Within-Subjects Contrasts table or the Tests of Within-Subjects Effects table. The former is shown below. You can find the main effect of "CASE" and "FREQUENCY" in the second and fifth row, and two-way and three-way interactions of the three variables of the study in the 3rd, 6th, 8th, and 9th rows. You may need the information in the columns of df, F, and Sig for reporting analysis results.

3c. The main effect of the between-subject variable:

The results of the between-subjects variable analysis can be found in the Tests of Between-Subjects Effects table.

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

You are not done yet!

For a typical psycholinguistic study involving a multivariable design and random selection of test material, you need to do the following:

1. Both subject analysis and item analysis.

2. Analysis of both reation times and error rates.

3. If there is a main effect or interaction, you also need to do post hoc analyses to determine, e.g., the differences found between two conditions for each participant group are significant.

Confused? Contact the Instructor.