MirageSF
asked on
Creating a random number generator with a defined start/end range and standard deviation
Hi,
I have a sheet that creates 5 cols by 50 rows (250 numbers) using randbetween of two defined values, usually between 5 and 30. These individual rows are then averaged (5 numbers averaged to generate 1) so in total 50 numbers.
These values are then used to create a normal distribution histogram.
However, I would like to define a number in row 1 of each of the 5 columns which will act as the standard deviation expected. So once the figures are created and the average of the 5 cols is created, the overall standard deviation of the 50 figures is equal to that defined.
How could I do this?
My initial sheet attached
I have a sheet that creates 5 cols by 50 rows (250 numbers) using randbetween of two defined values, usually between 5 and 30. These individual rows are then averaged (5 numbers averaged to generate 1) so in total 50 numbers.
These values are then used to create a normal distribution histogram.
However, I would like to define a number in row 1 of each of the 5 columns which will act as the standard deviation expected. So once the figures are created and the average of the 5 cols is created, the overall standard deviation of the 50 figures is equal to that defined.
How could I do this?
My initial sheet attached
MirageSFP,
Few questions, please...
(1) Is B1 the target? (The sheet says B10.)
(2) Did you mean the target to be 10? That's an enormous distance from anything that would be expected.
(3) "Close as possible" needs to be defined. +/- 10%, 1%, .1%?
Edit: What version of Excel?
Brian.
Few questions, please...
(1) Is B1 the target? (The sheet says B10.)
(2) Did you mean the target to be 10? That's an enormous distance from anything that would be expected.
(3) "Close as possible" needs to be defined. +/- 10%, 1%, .1%?
Edit: What version of Excel?
Brian.
MirageSFP,
Please see the attached. Press the blue button to run a macro that cycles through up to 100 random iterations looking for the first solution within 1% of the target. The code is...
Brian.Sample--4--V2.xlsm
Please see the attached. Press the blue button to run a macro that cycles through up to 100 random iterations looking for the first solution within 1% of the target. The code is...
Option Explicit
Sub Randomish()
Dim xMin As Double
Dim xActual As Double
Dim xTarget As Double
Dim i As Long
Sheets("Sheet2").Activate
xTarget = Range("B1")
xMin = 100000000
Application.ScreenUpdating = False
For i = 1 To 100
Range("B6:F105").Formula = "=(RANDBETWEEN($B$3,$B$4)+(RANDBETWEEN(0,9)/100))"
Range("B6:F105").Copy
Range("B6:F105").PasteSpecial xlPasteValues
xActual = Range("H1")
If Abs(xTarget - xActual) / xTarget < 0.01 Then
MsgBox ("Run complete - difference is " & Format(Abs(xTarget - xActual) / xTarget, "0.00%"))
Exit Sub
End If
If Abs(xTarget - xActual) < xMin Then xMin = Abs(xTarget - xActual)
Next
MsgBox ("Sorry - couldn't get within reach. Closest was " & Format(xMin / xTarget, "0.00%") & ". Please try again.")
Application.ScreenUpdating = True
End SubBrian.Sample--4--V2.xlsm
ASKER
Hi Brian,
Thanx for that, problem is obviously if Randbetween values are say 8-15 as they are then unlikely to get a 10 deviation which is expected, however even when I change the range from 8-30 and set the deviation to 10, it does not seem to get much higher than 3?
1) Yes sorry typo B1 is target.
2) Would not expect it to be bang on 10, but near as possible, expecting exactly 10 would cause massive delay in calculating.
3) I would say +/- 20% would be a exceptable range.
4) Excel 2010.
Thank you
Thanx for that, problem is obviously if Randbetween values are say 8-15 as they are then unlikely to get a 10 deviation which is expected, however even when I change the range from 8-30 and set the deviation to 10, it does not seem to get much higher than 3?
1) Yes sorry typo B1 is target.
2) Would not expect it to be bang on 10, but near as possible, expecting exactly 10 would cause massive delay in calculating.
3) I would say +/- 20% would be a exceptable range.
4) Excel 2010.
Thank you
MirageSF,
expecting exactly 10 would cause massive delay in calculating.
Spot on - an SD of 10 isn't quite age of the universe stuff, but I think it'd comfortably see the pair of us dead and buried. (For normally distributed data, the odds of getting an SD of even 7 is 1/390,682,215,445.)
What is the table for?
Regards,
Brian.
expecting exactly 10 would cause massive delay in calculating.
Spot on - an SD of 10 isn't quite age of the universe stuff, but I think it'd comfortably see the pair of us dead and buried. (For normally distributed data, the odds of getting an SD of even 7 is 1/390,682,215,445.)
What is the table for?
Regards,
Brian.
ASKER
Brian, like the dead and buried before calculation lol.
The table will be used for simulation data, generally looking for a deviation for each set of around 1-3 points, but occassionaly will need this to be around 10 if defined, although rare and more likely to stay 5 or below.
Is it not possible to have the random values as they are created now, but occassionaly add a bump value of say 10 to force the overall average up and therefore bring the deviation closer to target?
Cheers
The table will be used for simulation data, generally looking for a deviation for each set of around 1-3 points, but occassionaly will need this to be around 10 if defined, although rare and more likely to stay 5 or below.
Is it not possible to have the random values as they are created now, but occassionaly add a bump value of say 10 to force the overall average up and therefore bring the deviation closer to target?
Cheers
ASKER CERTIFIED SOLUTION
membership
This solution is only available to members.
To access this solution, you must be a member of Experts Exchange.
The_Barman,
Yes, that was my first thought, but this one is a lot simpler (assuming that MirageSF is content with the above).
Regards,
Brian.
Yes, that was my first thought, but this one is a lot simpler (assuming that MirageSF is content with the above).
Regards,
Brian.
ASKER
Hi Barman,
Yes you are correct, it was your mathematical genius which created that for me, for which I have been using for a while now. However, that version allowed me to define the range i.e. 8-15 which was great, but no control really over individual Average & Standard Deviation values which I required later on.
The updated version I am using NORM.DIST to generate my random numbers which gives me good control over average and standard deviation, but zero control over the range!. Unless I use MEDIAN in front of the NORM.DIST command as matthew Patrick pointed out, but this then does not create normal distribution.
So each column of 50 numbers needs range from range to, deviation and average expexted values, and then needs to generate 50 numbers with these factors which would be normally distributed.
Yes you are correct, it was your mathematical genius which created that for me, for which I have been using for a while now. However, that version allowed me to define the range i.e. 8-15 which was great, but no control really over individual Average & Standard Deviation values which I required later on.
The updated version I am using NORM.DIST to generate my random numbers which gives me good control over average and standard deviation, but zero control over the range!. Unless I use MEDIAN in front of the NORM.DIST command as matthew Patrick pointed out, but this then does not create normal distribution.
So each column of 50 numbers needs range from range to, deviation and average expexted values, and then needs to generate 50 numbers with these factors which would be normally distributed.
ASKER
Sample.xlsx