I have gathered quantitative data through content analysis and it is nominal data. I want to check the influence between two variables. So, Chi-square test of independence is appropriate for it? please suggest.
Yes, Chi-square would be the obvious way to evaluate the relationship between any kind of nominal or "count" data, regardless of its source.
With nominal data you can only test for co-occurrence or contingency. Chi-square seems reasonable if you have constant numbers of cases. If you want to compare values, a standardized coefficient, such as Cramer V would be useful. These coefficients do not show influence, however, but only co-occurrence.
If you have a temporal dimension, however, you can test whether the co-occurrence of elements increases or decreases and whether the occurrence of elements at an earlier time lead to the occurrence of other elements ar a later time. There, you can make some statements on influence.
As in all analyses, you can not find any form of influence or causality in cross-sectional analyses but need at least some temporal dimension so the cause may come before the effect.
Chi square is the best test to study association between two or more variables in a cross sectional nominal data. Please do take care to apply Yates correction wherever is the cell value in the table observed zero, because the resultant value reflects the data and the analysis is correlated through the observed frequencies.
best wishes
The level of association indicates some relationship exists between the variables but cannot be considered as causal factors because chi square is after all a non parametric correlational study. The result is analysed on the frequency of Data.
I suggest numerical coding for nominal variables:
https://www.researchgate.net/publication/278403191_Numerical_Coding_of_Nominal_Data
- Badges
- Science topic