How do I calculate a standard deviation of grouped data in Minitab? 
Request for Question Clarification by
elmartoga
on
18 Jul 2003 10:34 PDT
Hello len,
Just to make sure I'll give you the correct answer, does your
spreadsheet look something like this?
Year Sales
2000 100
2000 150
2000 130
2001 210
2001 250
2002 200
so the "groups" here would be 2001 and 2002. What you want to do is
calculate the standard deviation of elements in group 1 and the
standard deviation of elements in group 2?
Best regards,
elmarto

Clarification of Question by
_len_ga
on
18 Jul 2003 12:01 PDT
Thanks for the question. The grouping is based on a attribute rather
than a date. I'm not sure it makes any difference, but it'd probably
better if I explain it.
Suppose I'm manufacturing widgets every day. The time it takes to
make widgets is variable, though, for a variety of reasons we don't
care about. Suppose I measure the amount of time it takes, on
average, to make widgets each day, and it looks like this:
Widgets Seconds per
Day Made Widget
  
1 100 60
2 92 71
3 115 56
4 85 63
5 72 57
6 60 55
7 55 50
I don't know how long it takes, for example, to make any individual
widget on Day 1. All I know is that it took 6000 seconds (100 * 60)
to make 100 of them. I believe this is what makes the widgets
"grouped."
Using the number of widgets made on each day, plus the average time
per day, I can calculate the weighted average for making widgets for
the entire week. This is easy in Excel. I can also calculate in
Excel the standard deviation if I know the formula for calculating the
std dev of grouped data. My question is how I'd calculate the std dev
of this grouped data in Minitab.
Hope this helps. Thanks very much.
Len

Request for Question Clarification by
elmartoga
on
18 Jul 2003 19:37 PDT
Hi again, len!
OK, let me see if I understand. You're defining the widgets made in
day 1 as group 1, the widgets made in day 2 as group 2, etc, right? If
this is the case, then you can't calculate the std dev of each group
with the given data.
Let's take group 1. You only know that the average time is 60 seconds.
But since you don't know how much time each of the widgets took, you
can't calculate the std dev for this group. Same for the other groups.
If you wanted to calculate the weighted std dev across groups, I think
you would have to manipulate the columns in Excel, paste them in
Minitab and calculate the basic std dev. That is, you would have to
create a column with a hundred 60's, ninetytwo 71's, etc. I don't see
any other way to this (some other researcher may, though).
Please do tell me if I didn't understand correctly what you meant by
groups.
Best wishes!
elmarto

Clarification of Question by
_len_ga
on
19 Jul 2003 04:14 PDT
Hi elmarto,
Yes, I'd thought of doing a cutandpaste from Excel to Minitab with
one line per widget. The only problem is that I've got more than a
million widgets to measure, and Excel can only handle a fraction of
that.
I may end up sticking with my handmade calculations in Excel. I can
show how the calculations are derived, and I think that's more
important than seeing it done in a particular tool.
Len

Request for Question Clarification by
elmartoga
on
19 Jul 2003 09:28 PDT
Hi len!
Well, here's my last attempt. Have you tried the "Make Patterned Data"
feature in the "Calc" menu? If you do't have many observations (i.e.,
many days in the example), this might work to make one observation per
widget. So if you have one hundred 60's and seventy 80's, choose in
the "Make Patterned Data" dialog box:
Store patterned data in: some new column
From first value: 60
To last value: 60
In steps of: 1
List each value: 100 times
and the same for each day, always in a new column. Then you can use
the "Stack Columns" option in the "Manip" menu to make one long column
with all the new columns, and calculate the std dev of that column.
Please tell me if this worked you.
Best wishes!
elmarto
