I want to examine if there is a relationship between a continuous variable and a nominal one with 5 categories.
Actually the question is nonsense. It would only make some sense when the categorical variable was ordinal rather then nominal. If so, the biseral correlation coefficient would estimate the association.
You might also think in terms of models expressing one of the variable given the other. When the continuous variable is the response, a very simple (generalized) linear model would do. When the nominal variable is the response you'd need a multinomial logistic regression.
Andreas,
It depends somewhat on the hypotheses and the nature of the nominal variable. But it sounds as though it would be most easily modeled as either an ANOVA (random effects ANOVA, depending on the nominal variable) or a multinomial logistic regression.
Pat
Thank you both. My dependent variable is continuous. The independent is categorical (marital status coded 1-5). It is not ordinal. One way ANOVA is a good idea in order to compare the mean differences between the groups of the independent variable.
I am not sure, but I think that the point biserial correlation coefficient is used when the one variable is continuous and the other is binary/dichotomous (coded 1,0).
Andreas, yes: point-biseral for binary, and biseral for ordinal (with more than 2 levels/categories).
Note that ANOVA is an analysis of variances, not precisely ^what you want to do. But an ANOVA is one side-product or spin-off of a linear model. Further note that, depending on the *kind* of your continuous data (counts, proportions, concentrations, ...) it might be more appropriate to use a different error probability distribution than the normal distribution.
OK Jochen. Thank you. According to Andy Field (Discovering Statistics using SPSS, 3rd edition, p.182) the point-biserial correlation coefficient is used when one variable is a discrete dichotomy (e.g. pregnancy), whereas the biserial correlation coefficient is used when one variable is a continuous dichotomy (e.g. passing or failing an exam).
Maybe Spearman correlation could be a better choice instead of ANOVA due to the kind of my data. Or what do you believe about using Eta (Internal by nominal)?
there is a possibility, which is analyzed through the independence test for contingency tables Chi Square test. but would have to transform continuous variable to ordinal,
to do this, could you answer eg
does the use of certain types of soil, with the appearance of certain insect?
you could also use Cramer's V, but you should transform quantitative ordinal variable. although this procedure you can lose a bit of information, it helps to have notion of the trend.
If you are trying to test if there is a relationship between a continous variable (CONVAR ) and a 5 level nominal variable (CATVAR). I would use ANOVA CONVAR by CATVAR(1,5) or the Non Parmetric alternative Kruskall Wallis CONVAR by CATVAR(1,5),
The choice depends on the normality of the CONVAR and the equality of variance within each group.
My suggestion: http://www.researchgate.net/publication/278403191_Numerical_Coding_of_Nominal_Data
Article Numerical Coding of Nominal Data
- Badges
- Science topic