Quantitative Research Project

Mushy

Hey all, I be doing a quants reearch project for college at the moment. Just wondering could I have a tiny bit of help.

Am trying to find out what sort of variable certain things are(vote in last election, member of political party, interest in politics). Are they nominal or ordinal? I have a rough idea, but not sure. 1st and 3rd are ordinal anyway AFAIK. Trying to do the crosstabs stuff at moment, just wanna make sure they right.

EDIT: Also, include gender in that list. Thought it was ordinal, but gamma tests show no positive or negative. Think it must be nominal then.

morphem

Hey!
Look nominal scale is the one when you can find difference between variables. Nothing more than that. When you have gender:
1 - male or 1 - female
2 - female 2 - male
you can't measure that in any other way. It's not an ordinal scale.
Rest of stuff: vote in last election, member of political party, interest in politics depends on the cafeteria of answers. If an answer for the question of voting is 'yes' or 'no' it's again nominal level of analysis.

Write more details and your hypothesis, then I'll try to tell you what to do.

Ostrom

morphem wrote: »

Hey!
Look nominal scale is the one when you can find difference between variables. Nothing more than that. When you have gender:
1 - male or 1 - female
2 - female 2 - male
you can't measure that in any other way. It's not an ordinal scale.
Rest of stuff: vote in last election, member of political party, interest in politics depends on the cafeteria of answers. If an answer for the question of voting is 'yes' or 'no' it's again nominal level of analysis.

Write more details and your hypothesis, then I'll try to tell you what to do.

There are differences between all response categories. 'Level of measurement' is the mathematical precision with which you can differentiate between the categories.

Variables are typically referred to as either 'categorical' or 'quantitative' in sociology

Your three (really 4) levels of measurement are Nominal, Ordinal, and Interval

Nominal variables differ in kind, but not in quantity (religion, gender, race, ethnicity etc). There is no meaningful numerical separation between the values of the variables (catholic isn't one unit more religion than jewish etc...)

Ordinal variables can be rank ordered, but the distance between categories is imprecise - likert scales, statements such as 'higher, middle, lower; agree, strongly agree etc.... An order may be established, they differ in kind and magnitude, but it cannot be specified.

Interval is your highest level - 'real' numbers - age, income, time, distance. The distances between values are specific, they can be added, subtracted, multiplied and divided. Generally you should aim for the highest level wheer possible (in the social sciences you are limited to ordinal and nominal with most variables).

An ordinal measure of age would be:

15-21
22-26
27-30

An interval measure of age would be to ask the person their actual age. In the former, two distinct differing answers can be rank-ordered.

Vote yes-no is nominal, as is political affiliation. Measuring an individuals level of interest (depending on the measure used) is ordinal (higher, lower etc.)

Gamma is suitable for ordinal data, but you can also use chi-square (be careful with your cell counts). With interval measures you can use more powerful correlation statistics such as Pearson's R, but you can use Spearmann's rho on ordinal data.