Dear All,
I have these data. Two are continous and one are categorical. I wonder, is there any method to perform a correlation analysis on the impact of two continous data on one categorical data?
The example of the data is attached below
Hello Muhamad,
Unfortunately, I'm not clear as to what it is you seek. If you want to know how scores from sets of variables (which are continuous) relate, then there are:
1. Pearson correlation (pairs of variables at a time);
2. Multiple linear regression (a set of variables combined to yield the "best" possible correlation with a single dependent variable);
3. Multivariate regression (a set of multiple linear regressions);
4. Canonical correlation (like #3, an option for understanding how two sets of variables relate to one another);
5. Factor analysis (how the relationships among a set of variables helps to identify underlying structure among the set of variables, the structure implying the identification of one or more latent variables or factors);
...and others (e.g., Jacob Cohen's set correlation).
If you could articulate your specific research question(s), the specific variables and their quantification, and the data collection method, then I'd wager you would get a lot more focused recommendations.
Good luck with your work.
It looks like you have a dichotomous categorical variable (that is, it has only two levels). And it sounds like you want to treat the categorical variable as your dependent variable in the end.
First, it may be helpful to find the correlation between each pair of individual variables. For the categorical and one continuous, you can simply code the categorical variable as 0 and 1, and use a common correlation technique (Pearson, Spearman, Kendall). Probably the recommended method is to use point-biserial correlation. This is similar to what I just recommended and using Pearson correlation, but with some implementations there is a small difference due to how the standard deviation is calculated.
For the two continuous variables, use one of Pearson, Spearman, or Kendall correlation.
To determine the impact of the two continuous variables on the categorical variable, you can use multiple logistic regression. You would then probably want to use a goodness of fit measure, something akin to an r-square value for regression. I think what may make the most sense is to determine for how many cases your logistic regression equation will predict the actual value of the dichotomous variable. This approach is listed as "Count" here : https://stats.idre.ucla.edu/other/mult-pkg/faq/general/faq-what-are-pseudo-r-squareds/
Two options:
1. Point-Biserial Correlation: dichotomous variable (Visibility) vs continuous scale var (Arc L & Arc V). However, you need verify if your data (cont), are approximately normally distributed for each category of the dichotomous variable.
See this example (using R): https://www.statology.org/point-biserial-correlation-r/
2.- The non-parametric Mann-Whitney test, could also be another option.
- Badges
- Science method
- Similar topics
- Mathematics
- Statistics
- Statistical Modeling
- Analyses of Variance
- ANOVA