# How to analyze a 34 item standardised survey purportedly measuring 6 dimensions using SPSS?

I am currently analyzing a 34-item standardized survey on Physics expectations. I gathered data from 104 female respondents from a certain school. The survey's supporting literature said that the survey measures 6 specific dimensions, with each dimension corresponding to 5-6 items. My professor asked me to conduct SPSS analysis on the data aside from the usual mean and standard deviation. Now, I am in the dilemma of identifying the correct statistical analysis to apply using the SPSS.

Can anyone here please help me with this? What analyses can I apply?

This is a project for a Statistics class, and we have no specific questions to answer. We are just asked to make meaning from the gathered data out of various statistical methods. Thank you very much for the response.

I wrote a post a while back about some basic analyses that people often perform on correlational survey type datasets using SPSS. The outline of the post:

1. Get the raw data into SPSS
3. Check data
4. Compute Variables
5. Describing scale properties (reliability analysis and factor analysis)
6. Calculate univariate descriptive statistics
7. Obtain correlation matrix
8. Perform model testing

The post is designed for students in applied fields who are getting started with basic analyses using SPSS. It includes links to assorted SPSS tutorials.

In particular, given the description of your task, it sounds like factor analysis might be particularly relevant to you in order to check whether your data is consistent with the factor structure of the scale.

As for what model testing you might do, that would depend on what other variables you have in the dataset and what is theoretically interesting.

• thanks for this, Jeromy! If I do factor analysis, what measures can tell me that my data is consistent to the factor structure of the scale?
– Deah
Jun 17, 2012 at 7:29
• @Deah It sounds like you need to read more about factor analysis. At a really basic level, extract the specified number of factors in the test with a rotation and see whether the items have highest loadings on factors as specified by the test. There's a lot more you could do of course (e.g., look at tests of how many factors should be specified; look at confirmatory factor analysis; etc.). Jun 17, 2012 at 23:42
• Yes, we haven't discussed factor analysis yet and I have difficulty understanding it and using SPSS with it. Anyway, thanks a lot ! :)))
– Deah
Jun 18, 2012 at 0:51

For each domain the surveys are usually designed so that if you sum up the scores for all questions in that domain you have a meaningful measure of the domain's characteristic. Then you can compare the mean difference in the domain scores by group. A t test can usually be applied but you can always use the Wilcoxon rank sum test.

• Thank you! But based on my readings (correct me if I'm wrong), a t-test is a parametric test. I do not know how to check if my data has normal distribution.
– Deah
Jun 17, 2012 at 7:31
• Yes a t test is based on normality but is robust to mild departues from normality. Generally speaking survey data on a Likert scale is not going to even be approximately normal. Normality can be assessed using goodness of fit tests such as the Anderson-Darling or Shapiro-Wilk tests. When you sum several scores as you would do with domain comparisons the sum will be close to normal because of the central limit theorem. But the Wilcoxon rank sum test is a nonparametric test that can be used regardless and it does not require normality. Jun 17, 2012 at 12:13