I'm doing my research on modelling of barriers (limitations) and determinants (social constraints) of maize farmers' adoption of adaptation options to curb the negative impacts of CVC (climate variability and change) using the MNL model.
For the results of the MNL regression model, I'm using the assumptions of IIA (the Hausman test), the VIF, and the contingency coefficient that need deep result interpretation.
But in the STATA software, both the terms contingency coefficient and correlation coefficient appear, of course with different values.
What is the conceptual difference between the contingency coefficient and correlation coefficient? What best example do we have to differentiate them for better understanding?
THANK YOU SO MUCH!!!
A contingency coefficient is unaffected by the order of the categories of the variables. It gives a rough measure of the degree to which the rows and the columns of a table are associated. Cramer's contingency coefficient is based on the ratio of the observed Chi-squared to the maximum value it might have. I don't know that any of these coefficients have a real life interpretation, and would be interested to hear from anyone who is proposing one!
Correlation coefficients, on the other hand, imply an ordered association.
@Ronán Michael
Thank you for your professional supportive answer!
Still, I wonder why it is so-called "contingency"
Contingency = a future event or circumstance which is possible but cannot be predicted with certainty.
Contingent derives from the latin tangere to touch and con- meaning together, mutual. So contingency occurs when one thing influences another. A contingency, in the sense of a future circumstance that is possible but not certain, means an event that might happen subject to other events or conditions. So when you bring an umbrella you are preparing for the contingency of rain later.
Dear Ronán Michael Conroy
Really thanks a lot for your nice conceptual explanations!
STAY BLESSED!!!
A contingency coefficient is particularly informative if you’re working with a large sample, and you don’t need to find out if an association is complete or not—just whether or not the association exists for categorical variables.
A correlation coefficient is a measure of the strength of the relationship between two continuous variables.
Dear Teshita Uke Chikako
Really thank U so much.
...association...; complete or not— association exists; for categorical variables...impo words.
Can you please help me with a conceptual example? Contingency coefficient < or = 0.75. Correlation coefficient b/n -1 and +1.
STAY BLESSED!!!
Contingency in form of a contingency table talks about the association between two variables of data. Here both the variables are categorical and the measure is chi-square. When the variables are numeric we see a correlation. Here the measure is R square. Terms used are association and correlation.
Best!!
Dear Prof. Anuraj Nayarisseri , really thank you for your nice technical definitions.
Contingency is abt association of categorical variables; measured in X2
Correlation is abt correlation of numerical variables; measured in R2.
Hence, the concept lies in the understanding of the d/ce b/n association and correlation!!!! Am I right?
STAY BLESSED!!!
Alem
Alem Redda When you find relationship between two categorical variables it’s called association and when you find relationship between two numeric variables is called correlation. X^2 is chi-square and R^2 is R square.
Dear Prof. Anuraj Nayarisseri , thank you for making my modelling research less miserable!! You really clarified this so well.
So, the concern lies in the categorical vs numeric variables (of course, of the independent variables used in the model). For e.g. two categorical indep. variables used in the MNL model should have < or = 0.75 value; this means if it is > 0.75, they have a strong association and could substitute for each other, which does not give sense b/se their effect on the dependent variables would be the same. The indep. variables should stand alone.
STAY BLESSED for the effort you put into it.
Alem R
Dear Prof. Anuraj Nayarisseri , nice food for thought:
You reminded me of the saying: "To do big things, start from scratch".
The term "independent", well noted!!!
Thank U
Alem
Contingency and correlation are some of the most misunderstood and misused statistical terms used by researchers, the media, and the general public.
Here are the three major contingency and correlation pet peeves:
- 1) Correlation applies only to two numerical variables, such as height and weight. So, if you hear someone say, “It appears that the voting pattern is correlated with gender,” you know that’s statistically incorrect. Voting patterns and gender may be associated (i.e. contingency ), but they can’t be correlated in the statistical sense.
- 2) Correlation measures the strength and direction of a linear relationship. If the correlation is weak, you can say there is no linear relationship; however, some other type of relationship might exist, for example, a curve (such as supply and demand curves in economics).
- 3) Correlation doesn’t imply cause and effect. Suppose someone reports that the more people drink diet cola, the more weight they gain. If you’re a diet cola drinker, don’t panic just yet. This may be a freak of nature that someone stumbled onto. At most, it means more research needs to be done (for example, a well-designed experiment) to explore any possible connection.
Source:
Deborah Rumsey . 2010. Statistics essentials for dummies. Wiley Publishing, Inc. Indiana, USA.
- Badges
- Science topic
- Similar topics
- Correction