iSixSigma

Data type

Viewing 19 posts - 1 through 19 (of 19 total)
  • Author
    Posts
  • #32047

    YF Gao
    Member

    I have a basic statistic concept question afflicting me.
    According to different criteria: data can be catergorized as “attribute” and “variable” data, on the other hand, data can also be classified as “discrete” and “continuous” data.
    Some say: “attribute data” = “discrete data” and “variable data”=”continuous data”.
    Anyone can help in clarifying these concepts? Many thanks!

    0
    #85122

    Ed Van Haute
    Participant

    Attribute data has only two possible outcomes: yes/no, go/no go, etc.
    Variable data can be either discrete or continuous.
    Discrete data us measured in units, and is countable. # of people, 3 of items passing test, number of calls arriving, etc.
    Continuous data can assume any numerical value, and is dependent on the sensitivity of the measuring device.

    0
    #85123

    Ed Van Haute
    Participant

    Mis-keyed part of my reply. Part of the third line should read “# of items passing test”.

    0
    #85128

    Zilgo
    Member

    So if I had a marble factory and was producing colored marbles, say white, red, and black.  What kind of data would color be?  It is attribute.  Just because it has more than two possibilities doesn’t make it not attribute.  Attribute = Discrete, Variable = Continuous.  The difference between having two discrete categories versus more than two just means you no longer have binomial data.

    0
    #85138

    Ed Van Haute
    Participant

    Color is an example of “Qualitative” data. Attribute data by definition has only two possible outcomes. Discrete data is Quantitative, variable data that can be counted, and can have mathmatical operations performed on it.

    0
    #85141

    Zilgo
    Member

    So Qualitative data can’t have mathematical operations performed on it?  There are whole courses on analyzing qualitative data, and it can be both discrete and continuous.

    0
    #85143

    Ed Van Haute
    Participant

    Qualitative data, in it’s raw form, cannot by it’s nature, have mathmetical operations performed on it. To use your example, how do you add or multiply “red” and “black”. What you can do is perform operations on the frequencies of occurrences. Those frequencies of occurrences are variable data, quantitative by nature. Quantitative data can be either continuous or discrete.
    As to my statement that Attribute data can only assume one ot two outcomes, I refer you to the glossary of IMPLEMENTING SIX SIGMA (Breyfogle), and page 53 of the same text. If this is a matter of interpretation-so be it.
    I think our discussion is becoming pedantic. Out here.
     

    0
    #85150

    Zilgo
    Member

    I’ll agree…

    0
    #85152

    YF Gao
    Member

    After reading these threads, I am still confusing about it. Anyone there can give a clear explanation? Thanks

    0
    #85156

    Schuette
    Participant

    Maybe I can help you clear this up a little.
    Attribute data=discrete=count data
    Variable data= continuous=measured
    Attribute data catagorizes the data.  To use the marble example used earlier, you may have marbles in a jar that can be 3 colors (red, green, & black).  You can’t really do anything with this data except to count it (which makes it by definition discrete data).  It is also true that there are only 2 outcomes to attribute data.  For example:  You count the number of red marbles to be 10 in a jar of 100 marbles.  Your two outcomes are either a red marble or not a red marble (some other color).  Other examples of attribute data would include yield (pass/fail), # of defects on a part, etc.  There are equations that use the frequencies which you can use to statistically determine if you’ve made a significant change or not.  In general though, these are not as powerful or as extensive as the equations used for variable data.
    Variable data is always measured and the outcome (measurement) can be any numerical value.  Discrete data is finite and therefore cannot be considered continuous (infinite).
    In general, if all you can do is count the numbers of something, it’s attribute data.  If you measure it to get some value (e.g. time, pressure, length), then it’s variable data.

    0
    #85157

    YF Gao
    Member

    Thanks Jim:
    I understand that attribute data is “discrete data”. However, are all “discrete data” belonging to “attribute data”?
    can I say “attribute data” is one subset of “discrete data”?
    The same question to the variable data and “continuous data”. Is “continuous data” one subset of “variable data” or on the reverse? or they are interchangeable?

    0
    #85158

    Ankur
    Participant

    Hi
    i also feel that all atttribute data is discrete.so we can safely say that attribute data is subset of discrete data.

    0
    #85159

    Ankur
    Participant

    Discrete data us measured in units, and is countable. # of people, 3 of items passing test, number of calls arriving, etc.
    Continuous data can assume any numerical value, and is dependent on the sensitivity of the measuring device.

    0
    #85189

    Schuette
    Participant

    Attribute and discrete data are used interchangeably as are the terms variable and continuous.  In both cases, one is not a subset of the other.

    0
    #85224

    Markert
    Participant

    All random variables are defined on one of four scales of measurement: nominal, ordinal, interval, and scale.  Interval and scale are continous which means that in any arbitrarily small window of obervsation there exist uncountably infinite possible values. In quality, interval and scale data are referred to traditionally as variables data — one takes continous measurements on a variable.  Attribute is a quality term for types of data and has come to generally apply to any data that is not continuous.  The historic intent was that Attribute would apply to go/no go but its use has expanded over time to include all nominal and ordinal data.  Nominal data is a scale of measurement consisting of unordered categories (statisticians often call this categorial data) like the colors of M&M’s.  Ordinal data consists of ordered categories, there is a natural ordering to the categories and this type of data consists of ratings (like a Lichert scale) and rankings.  The term discrete and continous are used by statisticians to classify random  variables into two classes. Numeric observsations on discrete random  variables can only take on integer values while continous random variables can take on infinite values.  Discrete data consists of counts and proportions that are taken on nominal or ordinal random variables.  As an example, the binomial distribution applies to discrete random variables that consists of the counts from two categories (e.g., pass or fail).  The Poisson distribution applies to counts of defects per unit of observation and is the basis for DMPO calculations.  The normal distribution applies to random variables on a continous scale of measurements.  I hope this long winded discussion is not too confusing.  But, don’t confuse the four scales of measurements, with the two broad categories of numeric data that define probability distributions. Unless one can turn nominal and ordinal data into numeric data (discrete counts), then a statistical analysis would not be possible.

    0
    #85231

    Hariprasad
    Participant

    Variable Data(Continous Data): Measured ona Continous scale
    Discrete Data/Attribute Data: These are like countables or classification. This further can be classified as Ordinal, Nominal and Binary data
    Ordinal: Countables.. No. of Deaths, No. of defects etc
    Nominal: High/Medium/Low or Red/Yellow/Blue/Green etc
    Binary: Good/Bad, Pass/Fail

    0
    #85240

    Tierradentro
    Participant

    Phil-the four scales you described are do not preclude the use of statistics. An example is using Contingency Tables for Nominal data or Kruskal-Wallis Test on Ordinal data. Typically as the scale’s resolution decreases i.e. Ratio (you called Scale) ->Interval -> Ordinal -> Nominal, then the amount of data that is needed to determine significance increases. Reliance on the distributions such as Normal, Students t, etc. cannot be used directly with anything other than Ratio data. JMP software does a nice job in allowing the user to define the data as Continuous, Ordinal or Nominal and allowing the appropriate test for the data type.

    0
    #85287

    Sheetal
    Member

    Attribute data has clearly defined, two possible outcomes e.g YES/NO; HEAD/TAIL etcHowever, Variable data gives you the options to choose.  It can be divided into continuos and discrete data also.Example of Discrete data: Scoring a Composition on 1/2/3/4 or 5Example of Continuos data: Radius of a circle being 1.5 cm. 

    0
    #85288

    YF Gao
    Member

    Sheetal:
    Your understanding is similar to my original understanding. However, I was also told that “attribute=discrete” and “variable=continuous”.
    Any “guru” can give a confirmation on this confusing? related reference is highly appreciated.

    0
Viewing 19 posts - 1 through 19 (of 19 total)

The forum ‘General’ is closed to new topics and replies.