I know its really silly to ask such a question but I am in a fix and need proper elaboration.
For example I have a set of ungroup data consists of ages starting from 20 to 29. There are in total 30 observations which are recorded as 20, 22, 26, 29, 24, 22, 23, 25, 22, 26, 28, 27, 21, 25, 26, 23, 24, 27, 21, 25, 24, 23, 22, 28, 24, 23, 21, 21, 25 and 28. As I have calculated the range I come up with two results i.e. 29-20 = 9 but some said it should be 10. Now whatever the range is, if I need 3 class intervals for grouping this ungroup data I have found the class size to be 3 for each interval. For example 9/3 = 3 and 10/3 = 3.33 or 3 if I round it up to the closest whole number.
Now the main problem I face is to determine class intervals. If I plot it as: 20-23, 24-27 and 28-31 then it ends with 31 and some says that the class size is 4 here rather being 3. If I consider it as 20-22, 23-25 and 26-28 then it ends with 28 which have no option to record the value 29. If I develop it as 20-23, 23-26 and 26-29 again it come up with 4 class size and make me confused about placing an observation which as dual existence in the two class intervals, i.e. 23 and 26.
Can anyone explain what class intervals in must take which will have equal class size.
Hi, remember that two classes must be mutually exclusive in order not to contain a same value. So, for instance, you must decide whether the interval 21-23 includes or not the value 23 (that is, whether the interval is open or closed). This question is similar to making a histogram. The number of classes is somewhat arbitrary, but there are some rules of thumb to decide the number of intervals (see for example Sturges' rule).
Check also http://www.oxfordmathcenter.com/drupal7/node/255
Cheers
First of all: you claim that your data are ungrouped, but it is not so. Being 20 years old means that the time (in years) that has passed since your birth is in the interval [20,21[. Therefore you have observation in the interval [20,30[ , and this interval has length 10.
There are several compeding rules of thumb that are used to determine the "optimal" number class interval. What optimal should mean will depend on what you are going to use the intervals for. For histograms the most common rules are the square root rule that suggest that the number of intervals should be n^(1/2)=5.48 and Sturge's formula that suggests to use 1+ln(n)/ln(2)=5.91. I would round these numbers down to 5 so that I would take five intervals each of length 2 and starting at 20.
If you want perfectly equal class sizes, can you calculate age to more detail? For example, could you calculate someone's age as 22 years and 6 months = 22.5. If so, then you could just use class intervals of 3.33 in width.
Regardless, it really shouldn't matter. There's no right or wrong way to make intervals. There's no rule that says they have to be equal sizes, just whatever makes sense for the context of the problem. I often have age groups of varying size.
Mr Kabir.
You can use the Sturges method to determine the number of class intervals. considering only the integer part, divide the Range (max-min)by the numbers of intervals, to obtain the interclass range.
In the value of the interclass range, consider an additional decimal, the true value of the data of your variable. for example if the data of your variable is of type #. ##, then the interclass interval should be #. ### with the corresponding approximations.
in this way he obtains intervals of equal magnitude.
This is an objetive process
Aunque Sturges con muestras muy grandes pierde confiabilidad y precisión
bye
- Badges
- Science topic
- Similar topics
- Mathematics
- Statistics