In my case, positive integers, let's say a range of values between 1 and 100 is enough. The mean would then be 50. The SD is not important.

I also tested Excel's randbetween().

In cells A1:A1000 I put 'recordnumbers': Fill cell A1 with 1, fill A2 with A1+1, copy A2 to A3:A1000.
In cells B1:B1000 I put the RANDBETWEEN() function.
In cells C1:C1000 I put the RANDBETWEEN() function, but now 4 times: RANDBETWEEN()+RANDBETWEEN()+etc. (and then divided by 4).
In cell E1 I put the formula SMALL(B$1:B$1000;$A1) and this cell I copied to E2:F1000. It gives the ordered lists of columns B and C.
Then I selected E1:F1000 and chose a graph type, e.g. XY-scatter.

I've noticed that the E column ( one time RANDBETWEEN() ) gives a straight line, while the F column gives a curved line. The straight line indicates an even spread draw of values between 1-100. The curved line indicates that only few values are very low/high and most values are somewhere near 50.
However, I can't find any graph that shows me the distribution (the bell).

I've also tried it in Excel with values produced in VFP, using RAND(). In this case too the B column is straight and the C column is nicely curved.

But how can I prove that the values are really normally distributed (opposed to e.g. skewed)?? Is it the variance that must be of a certain value? I've forgotten.

As said before, the customer is satisfied if the values are SOMEWHAT normally distributed.


>Peter:
>
>What type of data do you want to populate this field? E.g. are they to be integers or real(floating) numbers or what? If non-integer, how many decimal places.
>
>What range of x values are required, e.g. -3 to +3? What range of y values, e.g. 1 to 5.
>
>If it is a standard normal distribution it will have a mean of 0 (i.e. be centred on zero) with a stand deviation of 1 (giving a measure of the spread of the distribution).
>
>I tested the Excel function randbetween() for an x range of -2 to 2 and took it out to 60,000 rows. This gives quite a rough standard normal distribution (the population mean is 0.00506859 and the population standard deviation is 1.153937829)
>
>The same parameters used with rand() gave a population mean of -0.006646157 and a population standard deviation of 1.156657729.
>
>The Excel randbetween() function looks to be smidgen less rough than rand().
>
>Heavy duty statistical packages like SPSS would have a much more refined ability to create standard normal distributions.
>
>Have you thought more about how this field is going to be used and what restrictions are placed upon it?
>
>Regards
>
>Godfrey