I fully agree with Alex. Any discretization of a continuous variable, however done, will lead of loss of information. But if it is really needed to be done, this must be driven by the question that needs to be answered, not arbitrarily.   

I agree with Ales and Subhash, but maybe Remilien wants to plot differences of age groups - then it's reasonable. cut() is your friend in this case.

Not sure to which languages/packages Gerhard is referring, but if using cut() in R it is

cut(X,k)

where X is your variable and k is the number of groups and it divides them to have approximately equal frequencies. So, to split 1000 drawn from the unit normal into 7 groups,

cut(rnorm(1000),7)

Of course often your criterion is not of that them equal in frequencies. For example, you might want your bins of equal length, then the many techniques for choosing bins in a histogram come into play.

Still, I agree with everyone that binning should only be done if you have a good reason for it.

See ?cut

The breaks can also be a numeric vector.

For this question i am offering only my personal opinion & absolutely not necessary to agree with the same.For our lifeline it is necessary to create the group that to for variable!!

In our life apart from our regular working formula & system we have to other important action for our life .Our life is a divine gift & we have to carry out our action which may be helpful to our family & to us personally which may grant us a fragrance in our life .

In this line instead of creating a group why we should not have a members /friends of our liking & test so that we may joyfully associated & also exchanged our line of thinking  which we may also help us in our career ,development & our growth sometimes it can be a messenger of our life .

I agree with the other posts. Binnig the data is not usually a good idea. It is also strange that it must be binned into four groups. So it sounds like what you are asking is given that I am making 4 bins and have a range from 24 to 81, how do I decide: is 24 to 30, and 30 to 56, and 56 to 75, and 75 to 81 just as good as anything else?

I have a bad feeling that you have data from people and insufficient data, but how about this: I have a range of ages from 24 to 81. Within these data there are four cohorts. The organism has one reproductive event per year, and many years are failures with no recruitment. On average, out to 81 years there will be four recruitment events to the population. So look at the data distribution and find these cohorts. I am guessing this is best done manually. Depending on the questions being asked, this might be a very appropriate application.

In general it is much better to find a more suitable analysis technique, and avoid binning the data. 

You may try with clustering with average linkage method for the single variable provided that the data available in advance.

First of all you have to check the frequency and percentage distribution. Accordingly then, you may break a variable into 2 categories using a cut point such as the median or you may break it into equal intervals. In sum, you can't tell unless you check how your data is distributed

You can also use hist() function with argument   plot=FALSE

Please see the help(hist) in R.

For whatever reasons you may decide to create groups from continuous variables, and what ever methods you decide to use (I should say there are several), it is important that you ensure that you have adequate frequencies in each group, so as to make for meaningful comparisons. For example, if after grouping, you have one group having, say 8 observations, and you have another group with 180 observations, you cannot make inferences based on comparisons resulting from groups with such wide variations. So my advise is: whatever drives your reasons for categorization, be sure that you have enough observations in the resulting groups, and be ready to justify your grouping criteria.