answersLogoWhite

0

This is a surprisingly difficult question, partly perhaps because of the ambiguous term 'ordinal'. For instance, horse-race finishes are ordinal--horses usually finish first, second, etc., with no ties; pure order data gives no information about gaps between horses. For such data--the purest form of ordinal data--talk about variability is meaningless. You need data with 'ties' or repeated 'values'--more cases than ordered categories--to talk about variability meaningfully.

If you do have repeated values, one option is to fall back and use nominal variability measures--the Index of Qualitative Variation is one; information statistics also work; and there's always the frequency/percentage table. They don't 'measure' concentration along the categoric order, obviously.

Disappointingly many websites recommend using the range or interquartile range, presumably calculated by assigning numbers to the ordered categories and subtracting. These indices are very dangerous if you assume only qualitative order among categories. This is obviously flawed--if you don't know how far categories are separated, subtracting numbers is flat invalid. For instance, rank states in the US by size--Alaska is 1, RI is 50--and consider the fact that a group from AK, TX, and CA has a range of 2 and a group from NJ and MA has range of 3 [47 - 44]. First, those numbers are really meaningless; second, they sure misrepresent relations among state size differences. Unless you trust that your 'ordinal' categories are pretty close to equal intervals apart--what we call 'quasi-interval'--you simply cannot use range validly to measure ordinal variability. The same reasoning applies to inter-quartile range. You might as well use variance, since describing 'skew' and 'outliers' for ordinal data is very dangerous, itself.

More valid ordinal measures do exist--I cannot recall them. But when you choose an index, take care to examine how it is treating the numbers or other ordering symbols it trades on. Invalidity is rife.

User Avatar

Wiki User

11y ago

Still curious? Ask our experts.

Chat with our AI personalities

ReneRene
Change my mind. I dare you.
Chat with Rene
JudyJudy
Simplicity is my specialty.
Chat with Judy
RossRoss
Every question is just a happy little opportunity.
Chat with Ross

Add your answer:

Earn +20 pts
Q: What is the appropriate measure of variability for ordinal data?
Write your answer...
Submit
Still have questions?
magnify glass
imp
Continue Learning about Statistics

IS GPA nominal data or ordinal?

It is ordinal.


Should The mean or median be used in ordinal data?

The median is used when reporting ordinal data.


What measure of central tendency may not exist for all numeric data sets?

Measurement Scale Best measure of the 'middle' Numerical mode Ordinal Median Interval Symmetrical data- mean skewed data median Ratio Symmetrical data- Mean skewed data median


What is the measure of central tendency for nominal and ordinal data?

The mode can be used with both kinds of data. The median may be used with ordinal data but great care is required if the median falls between two classes of observations.The mode can be used with both kinds of data. The median may be used with ordinal data but great care is required if the median falls between two classes of observations.The mode can be used with both kinds of data. The median may be used with ordinal data but great care is required if the median falls between two classes of observations.The mode can be used with both kinds of data. The median may be used with ordinal data but great care is required if the median falls between two classes of observations.


How do you convert ordered data to nominal data?

I think you mean ordinal data. Similar to the golf tournament, you need to determine where to "cut" (from the ordinal data) so as to divide the data into different categories (to the nominal data). For example, if the ordinal data range from 1 to 6 (where 1 = the best) and the cut is 3, then you convert all the numbers from 1 to 3 to "1" (which represents "good") and the all numbers from 4 to 6 to "2" (which represents "bad"). In other words, 1, 2, and 3 from the original ordinal data set are converted to "1" (ordinal data); whereas 4, 5, and 6 from the original date set now become "2" (ordinal data). Eddie T.C. Lam