I use this dataset. This is taken form classroom observation score on three point scale form 1 to 3. My independent variable is status. I want to know that;
1. Is this total and each construct is normally distributed?
2. Can I apply the t-test to find the significance between Status (1 & 2)?
For a real dataset you will never really know. As you have discrete numbers your data can at most be approximatly normal distributed. Instead of a t-test you could go for a corresponding rank test not requireing the normal assumption such as Wilcoxon-Mann-Whitney test.
hth
Matthias
Hello Mushtaq!
In SPSS you should to verify normality. for this you can follow many ways such as skewness and curtosis, histogram chart , box plot, levene tests (kolmogrov and shapiro) etc. if you have 30 or more samples you check kolmogrov-smirnov test results, on the other hand if your samples under 30 or low check for shapiro-wilk.
First, enter your data in SPSS, after follow step which below
Analyze>Descriptive Statistics>Explore
Move to "group" factor list and data moves to dependent list. and then click the "Plots" ,choose the "histogram and normality plots with tests" and than click OK.
Evaluate your normality results and choose parametric or nonparametric tests from SPSS.
I wish this helps you.
You can simply obtain the value of the skewness in SPSS, this would guide to know if it is positively skewed, approximately normal, or positively skewed. Better still, you can perform normality tests like the Kolmogorov Smirnov test, Anderson Darling Test, or Shapiro Wilk test. You can also plot the histogram for the data and include the normality curve. I hope this will guide you a little.
Pelumi, Yasin, Bruno, Matthias
This was my first experience to post a question. These are more than helpful. Thanks a lot every one.
Just to add to Yasin's point, the best normality test for small sample sizes like yours is Shapiro-Wilk rather than Kolmogorov-Smirnov. Even if your data is not normally distributed you can still run a t-test provided your sample sizes are equal and larger than 25 - see:
Sawilowsky, S. S. and Blair, R. C. (1992) A more realistic look at the robustness and Type II error properties of the t test to departures from population normality, Psychological Bulletin, 111(2), pp. 352–360.
- Badges
- Science topic
- Similar topics
- Mathematical Sciences
- Statistics
- Statistical Software
- SPSS