[DATA ANALYSIS] Nonparametric tests, partial correlation test, correlation tests and t-test in practice with SPSS

#In Data Analyst series

1. If there is any relationship between the age and income of the participants

Check normality: Both variables are normally distributed

Check linearity: Seems like there’s a linear relationship between age and income of the participants.

We will then run Pearson correlation analysis on these 2 variables.

Correlations

These 2 variables indeed correlate with each other with linear relationship.

2. If there is a difference between age and job satisfaction of the participants

Recode the age variable into 2 group: 1-from 0 till 35, 2-the rest ages.

Check normality for each group: OK

Independent samples t-test is used then.

Check variance homogeneity: p-value > 0.05, OK

Testing result:

p-value for both one-sided p and both-sided p <.05, reject the null hypothesis, meaning that job satisfaction does differ greatly for age group 0–35 and 35 up.

3. There is a difference between the income and management style of the participants

Check normality: OK

Recode the income into 2 groups: 1–0 till 12, 2-the rest.

Check normality for both groups: OK

Independent samples t-test is used.

Check variance homogeneity: OK

Testing result: p-value for both one-sided p and both-sided p <.05, reject the null hypothesis, meaning that management style does differ greatly for income.

4. We control the management styles, is there a relationship between job satisfaction and income

Check normality: OK

Partial correlation is used.

Result: While controlling management style, income and job satisfaction does correlate to each other. (p-values <.05)

5. Create your artificial pre and post-test results, and then be curious if:

· Your pretest scores differ based on the income

· Your post-test results differ based on the income

· There is a significant difference between pre and post-test results?

Create artificial pre-test and post-test results: (Transform -> Compute Variable -> random numbers -> R.Uniform(0,100)

Check normality: ok for skewness and kurtosis but fail the Normality test.

Consider this, I would choose to run nonparametric test for pretest and posttest scores to find if there’s any significant difference between these 2 variables.

Well, no they are not correlated due to p-value >.05.

Let’s then check if these scores differ based on income. We would still use the income_group variable that has been splitted above based on income variable.

Check normality: OK for skewness and kurtosis, pretest and posttest with income group 2 are normally distributed, but failed largely with group 1.

Therefore, for pretestStress with 2 income groups, we would use the independent samples t-test, while posttestStress with 2 income groups, we would use the independent samples t-test and nonparametric test.

Pre-test: failed the variance homogeneity test, we then have to then switch to nonparametric test.

Pre-test nonparametric test:

Reject the null hypothesis, meaning that the pre-test scores differ based on income group (below 12 and above 12).

Post-test: Again, failed the variance homogeneity test, switch to nonparametric test

Post-test nonparametric test:

Retain the hypothesis, meaning the post-test scores do not differ based on income group (above and below 12).