Hi,

You have choose to analyse your data by using ordinal regression, part of the process involves checking to make sure that the data you want to analyse can actually be analysed using ordinal regression. Your data, ie dependent data should be measured at the ordinal level.

Examples of ordinal variables include Likert items.

a.  Your Likert scale items: 7-point scale from ‘strongly agree’ through to ‘strongly disagree’, amongst other ways of ranking categories.

b.  3-point scale explaining how much a customer liked a product, ranging from ‘Not very much’, to ‘It is OK’, to ‘Yes, a lot’.

to use or not to use linear regression is not a matter of 7 point (or 5 point or 10 point) items but in the first place: linear relationship! :-)

then, of course, it's also a matter of variable type. in this case I don't know anything about your 7 point scale. were there verbal expressions at each point of the scale (like in Senthilvels example b - this is clearly ordinal) or were only the two endpoints labeled? in the last case most researchers argue that you can treat this as interval scale.

ordinal regression if you independent variables are qualified for it. 

Good article!!

Thank you all very much! your feedback was tremendously helpful

Hello Stav,

As all or friends suggested your objective analyses, the number of variables you want to have, number of items you have to measure a single variable, and several other factors determine the right regression you have to conduct.

IF it can be of help for you, I would like you also to consider linear regression provided that all the assumptions are met. As well my suggestion, will also require you of an additional assumption of factor analyses.

If you are confident that you can treat the dependent variable as a continuous scale provided that you have many items to measure it, you can use linear regression model. Tis is practicable when you want to treat that variable as a single variable and when you do not want to loose the power of analyses. In such case, you can conduct factor analyses provided that all the assumptions are fulfilled. The factor analyses will create regression score for you, which you will use as a dependent variable.These factor scores, however, are not actual scores. They are good in indicating you the correlation.association betwen variables when you use them in regression anlyses.

Regards

You need to consider the frequency distribution within each category for a reliable estimate. It is better to merge categories if some have disproportionally many or few frequencies.