In SPSS when values from external sources are read or a new variable is created and it has at least 24 unique values it is automatically classified as 'interval' (see http://127.0.0.1:60073/help/index.jsp?topic=%2Fcom.ibm.spss.statistics.help%2Fspss%2Fbase%2Fidh_pref_data.htm). If my memory service me correctly I had read somewhere that if an ordinal variable has more then 15 unique categories it is almost indistinguishable from an interval variable (if some could give me the references to the/an article that has investigated this, I would be happy (it might have been old help-functions within SPSS)). But why does SPSS takes this value as cut-off by default? Is there a cut-off where one can not distiguish parameter estimates in for instance a regression or SEM analysis between an ordinal and an interval variable?
BTW I have been trying to find the or a reference, but could not find it.
If a Likert scales has 7 steps or more you can consider it an interval scale according to teaching in Finland. ref.: Tabachnik & Fidell
sec.ed. using multivariate statistics, page 7.
Ordinal scales provide good information about the order of choices, such as in a customer satisfaction survey. Interval scales give us the order of values + the ability to quantify the difference between each one.
FOR MORE DETAILS, FOLLOW THIS LINK
hpstt://www.mymarketresearchmethods.com/types-of-data-nominal-ordinal-interval-ratio/
@Lakhwinder, That is very non informative. Maybe I should clarify. In IRT models sum scores on ordinal data for Rasch and Mokken models can be sufficient and necessary conditions, but with a Rasch model the level can be upgraded from ordinal to interval. However, in both models the total amount of measurement points remain the same after a correct model has been identified. 5 items with 4 categories will result in 16 distinct values. In Mokken they are ordinal, because the measurement requirements are not as strict as in Rasch. SPSS takes 24 distinct measurement points as the cut off between interval and ordinal. My question is how many distinct ordinal datapoints are required so one can not distinguish between ordinal and interval in the parameters in for instance regression analysis. In Nelder and McCulllugh (1st and 2nd) edition it is stated that statistics is about finding patterns in data. In this view, regression can be considered a short-hand form writing down an equation when data is interval.
- Badges
- Science method
- Similar topics
- Mathematics
- Statistics
- Statistical Techniques
- Structural Equation Modeling