Hello everyone. I have 2 sets of questionnaires with the following scale choice:
(1) Poor - Excellent
(2) Seldom - Always
Both are 5-point Likert scale.
The first scale has 10 items, asking for the comfortability of the employees in their workplace, they'll just rate it from poor to excellent.
While the second scale is measuring individual work performance with them answering how rare/often they do such things listed. There are 18 items in total.
I am having a hard time what to do, and yet my goal is to see a correlation between them. Thanks in advance for those who will help!
What you have are two multi-item scales. The first step its to assess the reliability of each scale using coefficient alpha, which requires all the items in each scale be scored in the same "directions" so that they are all positively correlated.
If you get an adequate alpha of .8 or above, the you can add the items on each scale together and do a correlation between them. Alternatively, if the alpha on either scale is low, the you should do an Exploratory Factor Analysis (EFA) to determine whether that set of items contains two or more subscales.
For your EFA, you should use either Principal Factors or Maximum Likelihood as your extraction method and specify correlated factors (an oblique rotation).
You may use this method:
Calculate the total or average scores for each set of questions for each respondent. This simplifies the data into two variables per respondent: Comfortability Score: Sum or average of the 10 comfortability items.
Performance Score: Sum or average of the 18 performance items.
Then check normality and choose the Appropriate Correlation Method: Pearson Correlation: Use if both summed scores are approximately normally distributed.
Spearman's Rank Correlation: Use if the data are not normally distributed or if the Likert scales are treated as ordinal data.
Then run correlation. By SPSS, Analyze > Correlate > Bivariate. Select the two new variables (comfort_score and performance_score). Choose Pearson or Spearman based on your data distribution. Click OK to run the analysis.
Heba Ramadan , this is a good way and I have added few other steps to correlate two sets of 5-point Likert scale questions:
1. Compute Mean Scores: Calculate the mean score for each set of Likert items for each respondent.
2. Check Assumptions: Ensure that data meet the assumptions for correlation (normality, linearity).
3. Use Spearman's Rank Correlation: If the data are ordinal or not normally distributed, use Spearman's rank correlation coefficient.
4. Use Pearson's Correlation: If the data approximate normal distribution, use Pearson’s correlation coefficient.
5. Statistical Software: Utilize statistical software like SPSS, R, or Python to compute the correlation.
Example in R:
```R
# Assuming you have two data frames: comfort (10 items) and performance (18 items)
comfort_mean
If you do not begin by assessing whether your items from a single scale (i.e., high reliability), you run the risk of adding together things that do not actually belong together. That is why you need to calculate coefficient alpha (and if necessary, exploratory factor analysis) prior to adding the items together.
Good day, sir David L Morgan Muhammad Haroon Shoukat Heba Ramadan
Thank you all for your helpful responses!
- Badges
- Science topic
- Similar topics
- Instrumentation Engineering
- Instruments