Timeline for Calculate mean of ordinal variable
Current License: CC BY-SA 3.0
21 events
| when toggle format | what | by | license | comment | |
|---|---|---|---|---|---|
| Aug 16, 2013 at 15:17 | comment | added | Nick Cox | I don't know how I can make it any clearer, but (e.g.) "0-4", "5-19", "20-114" are ordered (ordinal) in that there is only one natural order to those measurements (short of reversal). If you want to call them other things too, that's fine by me. | |
| Aug 16, 2013 at 12:48 | comment | added | SAAN | @TooTone "1="microscopic",2="tiny"" than interpret 1.5 | |
| Aug 16, 2013 at 12:43 | comment | added | TooTone | I can't add much to @NickCox's exposition. I was trying to think of an example where the categories are obviously skewed, so taking the mean of equally spaced category labels does not correspond to what the categories are labelling. "Splitting a numeric scale" as Nick said seemed a clear way of doing it. But you could also come up with something more subjective. E.g. weights of animals: 1="microscopic", 2="tiny", etc. | |
| Aug 16, 2013 at 11:52 | comment | added | SAAN | I think we are going on other track, @NickCox I need explanation why you are calling interval split data on real line as a ordinal data? | |
| Aug 16, 2013 at 11:39 | comment | added | Nick Cox | I think you can strengthen your answer and I encourage you do that. "because the mean might be an undefined value" is not a strong argument here, logically or psychologically, and does not focus on the deeper issue of whether equal differences really mean equal differences. | |
| Aug 16, 2013 at 11:31 | comment | added | SAAN | But in qualitative data (nominal and ordinal) we use percentages or ratio and in quantitative (interval and ratio) used averages. Mean is defined when sum divisible by 4, in fact its rare in all possible combination and we cant say this time defined and this time not. (we are not violating basic) | |
| Aug 16, 2013 at 11:21 | comment | added | Nick Cox | I agree; but my point was that the mean not being a possible value is not the key objection, as that is easily possible with variables you don't call ordinal. (Also, the mean could be a possible value with ordinal data, as with 1,2,2,3.) | |
| Aug 16, 2013 at 11:18 | comment | added | SAAN | @NickCox ok, In your last example "people in a family" has meaning full difference this characteristic belong to interval and also see the Glen_b comment on the question. | |
| Aug 16, 2013 at 11:05 | comment | added | Nick Cox | I don't think this is difficult. You could ask someone to report age category for people in a family, in years, 0-4, 5-9, etc. Sure, underlying that is age in integer years, and underlying that in turn is a continuous scale. But what the researcher sees and can analyse are in this example just ordinal categories. That's what you name, what the researcher has, not what it might be. | |
| Aug 16, 2013 at 11:00 | comment | added | SAAN | @NickCox Your answer is nice I cant say any thing about your answer, Yes my disagreement is with TooTone because in my knowledge the point (split a numeric scale) relates to mixed (ordinal and interval) not pure ordinal. | |
| Aug 16, 2013 at 10:54 | comment | added | Nick Cox | I didn't say anything about weight. Please look again: my example is children per family, who you count. Your disagreement is with @TooTone, who makes a perfectly valid point that you could split a numeric scale into ordered (ordinal) categories. | |
| Aug 16, 2013 at 10:46 | comment | added | SAAN | @NickCox "Weight of children" is quantitative measure and mean is appropriate for that, "number of marbles in a bag" I dont understand how it relates with the discussion and "something like small="0-1"..." is assumed infinite values between "0-1", "2-4" etc. | |
| Aug 16, 2013 at 10:38 | comment | added | Nick Cox | Azeem: I disagree. @Tootone has a good example; his example and yours are both ordinal. What Likert [NB] scales are is best left as a different issue. | |
| Aug 16, 2013 at 10:21 | comment | added | SAAN | @TooTone In fact Qualitative data based on nominal and ordinal classification and example you provide belongs to interval or Liket scale you are confusing your self in ordinal and Liket scale data. | |
| Aug 16, 2013 at 10:18 | comment | added | SAAN | @luciano How 2.5 mean make sense? You say halfway between disagree and agree (distance is 0.5) and same half way of strongly agree and strongly disagree (distance is 1.5). Ever you seen this? and how it possible one mean has two interpretation. Actually 2.5 is not representative part of your data, your data jumps integer to integer ranges 1 to 4. | |
| Aug 16, 2013 at 9:45 | comment | added | TooTone | @Azeem I like your approach of providing a simple counter-example. How about an example where your categories are quantifying something, e.g. weights of children, number of marbles in a bag, e.g. something like small = "0-1", medium="2-4", large = "5-10", xl = ">10"? | |
| Aug 16, 2013 at 9:30 | comment | added | Nick Cox | I think Azeem needs a stronger example. You could object to 2.5 as the average of 1, 2, 3, 4 children per family on the same grounds, how is that to be interpreted as it is not one of the defined values. That raises different issues. | |
| Aug 16, 2013 at 9:29 | comment | added | luciano | Agree mean of 2.5 in this context still makes sense to me - halfway between disagree and agree, or in other words, neutral. | |
| Aug 16, 2013 at 9:22 | comment | added | TooTone | Playing devil's advocate a little, in this example, I would interpret 2.5 as being half-way between 2, "disagree", and 3, "agree". This makes sense as an average given that we have "strongly disagree" vs "strongly agree", and "disagree" vs "agree". | |
| Aug 16, 2013 at 9:05 | history | edited | Nick Cox | CC BY-SA 3.0 |
added 26 characters in body
|
| Aug 16, 2013 at 8:59 | history | answered | SAAN | CC BY-SA 3.0 |