I want to see the association between nominal variable (dichotomous) and an ordinal variable, So can I use Chi-square test or some other test?
In this case, I believe that the test described by Mann-Whitney is more appropriate and that it consists of comparing each individual of the first group with each individual of the second group, registering many times favored in that comparison. Based on this count, to measure is constructed that is contrasted to see if the difference with the expected result, if there are differences between the groups, may or may not be attributed to chance.
I second what Angel has already said: A Chi-Squared test for Contingency tables will be fine.
If you want to do more, you may want to look up for Ordinal Regression Models. These are useful when you want to say more with the data than just say there is an association.
Hi, Yes you can but when you are analyzing the association for a R*C table (for xample a 3*4 ) using Chi square, your expected count should be lees than 20%. This is reported under your tables in SPSS.
If you use an ordinary Pearson chi-square, or the likelihood ratio chi-square, you will be treating the ordinal variable as nominal. With one dichotomous and one ordinal variable, you can use the "linear-by-linear association" test that you see in your SPSS CROSSTABS output. For more info, see Dave Howell's nice notes (link given below).
HTH.
https://www.uvm.edu/~dhowell/StatPages/More_Stuff/OrdinalChisq/OrdinalChiSq.html
Thank you everyone for your suggestions. All your guidance helped me in carrying out analysis.
There are possible several methods, for example one as attached below. But, Chi-square to the best of my knowledge provides information of level of significance or confidence in the association of two variables. However, it is phi that is measure of association and can be converted into chi-square to obtain level of significance. You can do reverse of it, if you have calculated chi-square value, convert it into phi that will give you measure or strength of association.
Everything sent by profesor mohammad Firoz Khan is a spectacular presentation of power point and I think that is enough to your problem
erick
It depends on how many values has the ordinal variable. If not many, and there are fulfilled assumptions - you can
Angel, how want you use Spearman's correlation in this situation? I think this is not a good idea.
CHI sqiarre test is a relational test between two varaibles in quantitative research. both variables have to be quantified in order to be corelated if one is not quantitief statastically it is impossible to corelate or find a relationship. that is my understanding. if that is not basiccaly done you with never related
thanks
In this case, I believe that the test described by Mann-Whitney is more appropriate and that it consists of comparing each individual of the first group with each individual of the second group, registering many times favored in that comparison. Based on this count, to measure is constructed that is contrasted to see if the difference with the expected result, if there are differences between the groups, may or may not be attributed to chance.
If you have ordinal independent variable and nominal dependent variable,
I think you can try Cochran-Armitage Trend Test. It would be helpful to check the trend of between two variables.
Kindly go through these links..
https://stats.stackexchange.com/questions/23938/how-to-correlate-ordinal-and-nominal-variables-in-spss
http://users.stat.ufl.edu/~aa/articles/agresti_1981.pdf
- Badges
- Science topic
- Similar topics
- Mathematical Sciences
- Statistics
- Statistical Software
- SPSS