I want to examine the direct relationships between workplace spirituality and employee attitudes (job satisfaction and organizational commitment) through correlation analyses. The Shapiro wilk tests for the questionnaire items of all three variables are significant. On the other hand, skeweness and kurtosis values are within the range +/- 1.2. Should I go for linear regression analysis?
Your data distribution may not be the only problem here I am afraid. The scales you are correlating are subjective ones and we would not usually consider it appropriate to use parametric tests on them. This is because it is not normally, no pun intended, thought wise to claim that people making judgements on such scales do so on a properly linear basis. The data therefore isn't thought to be ratio or even interval necessarily but more likely ordinal, so non-parametric tests are more conservative . See Sidney Siegel, non parametric statistics for the behavioural sciences if you need more information on this. I might be able to help further if you described exactly how the response scales are set up and the responses elicited.
Hi Arjun,
From your question, not sure your dependent variable's (DV) type is ordinal or interval / ratio scale and sample size collected. When your DV is in ordinal scale e.g. Likert scale, you should use non-parametric statistics. When your DV is in interval / ratio scale, you can use parametric statistics. Some scholars still debating Likert scale can be treated as interval scale & hence handled by parametric statistics e.g. Lubke & Muthen (2004). However, some argue the other way round i.e. Likert scale should use non-parametric statistics e.g. Jamieson (2004). In case your DV is in Likert scale & based on common observations, you can change your Likert to interval scale data through modifying the survey questionnaire before data collection i.e. converting Likert scale questionnaire (e.g. 1=Effective, 2=Some How Effective, 3=Moderately Effective, 4=Neutral, 5=Moderately Not Effective, 6=Some How Not Effective, 7=Not Effective) to Semantic Differential scale using Bipolar Adjectives (e.g. Effective1, 2, 3, 4, 5, 6, 7 Not Effective) so that parametric tests can be applied. The rationale claimed is that the intervals between the scale values can be treated as equal intervals, making it an interval scale which justified for parametric tests.
If your DV is in interval scale & assuming your sample size is < 50 since you are using Shapiro-Wilk test. May be you need to increase your random sample size & retest with shapiro-Wilk or even Kolmogorov-Smirnov when sample size > 100. If your p > 0.05 i.e. not significant then you have normal distribution for your sample data collected in which you can use parametric statistics. All the best.
Regards,
Fung
No, the data you will generate is likely to be on the ordinal scale so you cannot do linear regression. You can however use the spearman rank correlation,a non parametric test as opposed to the parametric test that you are considering .Best wishes
In general, parametric tests are robust to violations of normality, so you may be able to use parametric tests with your data even if the data are ordinal. This is particularly true if you are summing responses on Likert-type items. See Norman (2010) Likert scales, levels of measurement and the "laws" of statistics for details. There is nothing wrong with applying parametric tests to "subjective" data, and "subjective" doesn't equal "ordinal." I think with skewness and kurtosis measures within +/- 2, your data might be "normal enough." Tests of normality can be significant with even trivial departures from normality with large enough sample sizes.
Thanks to all of you for taking time to respond to my question. I would like to add so as to make my situation clear to you all. The three variables workplace spirituality, job satisfaction and organizational commitment were captured through questionnaire items (7-point likert scale). Workplace spirituality, job satisfaction and organizational commitment scales have 22, 3, and 8 items respectively. The sample size is 211. As per Sean's suggestion, I went through Norman's (2010) article. On the basis of this article, even though the distribution of the three variables is not normal, now I'm going to use parametric tests.
Thanks once again.
- Badges
- Science topic
- Similar topics
- Psychology