SPSS相关性检验及两时点调查数据分析求助（统计入门课）

Hi Eva, let’s walk through your SPSS correlation questions clearly—these are classic scenarios for stats intro courses, so I’ll break them down with actionable steps.

First, let’s match the right test to your variable types:

• One yes/no (dichotomous) variable + one Likert scale (ordinal) variable:
  • Start with Spearman’s Rho—it’s built for ordinal data (like Likert scales) and works seamlessly with dichotomous variables. If your yes/no variable is a true natural split (e.g., "did you participate?" vs. an artificial cutoff), you could also use Point-Biserial Correlation, but Spearman is more flexible for intro-level work.
  • SPSS steps: Analyze → Correlate → Bivariate, add both variables to the "Variables" box, check "Spearman" (or "Point-Biserial" if applicable), then click "OK".
• Two Likert scale variables:
  • Stick with Spearman’s Rho. Pearson’s Correlation is for continuous variables, and while some folks treat Likert scales as pseudo-continuous, Spearman is the safer, standard choice for ordinal data in intro stats.
• Two yes/no variables:
  • Use the Phi Coefficient—it’s tailored specifically for 2x2 categorical variable pairs. You can also run Spearman’s Rho here, but Phi is more precise for this exact case.
  • SPSS steps: Either use Analyze → Correlate → Bivariate and check "Phi", or go to Analyze → Descriptive Statistics → Crosstabs, add the variables to "Row" and "Column", click "Statistics", check "Phi and Cramer’s V", then "OK".

Let’s tackle each part of your dataset one by one:

• Same 10 questions at Time 1 vs. Time 2:
  • For each paired question, use the test matching its type:
    • If the questions are Likert/ordinal: Spearman’s Rho (same as above) to measure how consistent responses are across the two time points.
    • If the questions are yes/no: Phi Coefficient (for 2x2 paired categorical data) or Spearman’s Rho.
• Time 1 vs. Time 2 discussion frequency (4-level ordinal variable):
  • This is two paired ordinal variables, so Spearman’s Rho is perfect here. It will tell you how strongly someone’s discussion frequency at Time 1 correlates with their frequency at Time 2.
  • SPSS steps: Analyze → Correlate → Bivariate, select both discussion frequency variables, check "Spearman", then "OK".
• "Same location at both time points" (dichotomous variable) correlations:
  • If you want to see how this variable correlates with, say, Time 1 discussion frequency (ordinal), use Spearman’s Rho or Point-Biserial Correlation.
  • If you want to compare correlation strength between the "same location" and "different location" groups (e.g., does consistency in discussion frequency differ by location?), split the dataset first: Data → Split File → Organize output by groups, select the location variable, then run the Time 1/Time 2 correlation for each group separately.

A quick pro tip: Double-check that your variables are labeled as "Ordinal" or "Nominal" in SPSS’s Variable View—this helps avoid unexpected test results!

内容的提问来源于stack exchange，提问作者Eva