8284_zh-TW

liliconcom

Combining multiple std.devs or combining different datas<meta http-equiv="Content-Type" content="text/html; charset=UTF-8" />

Hi there!

I've created a questionnaire with three "categories" (Cx) and four sub questions (Qy) for each category:

C1:

C1Q1... (1-7)

C1Q2... (1-7)

C1Q3... (1-7)

C1Q4... (1-7)

C2:

C2Q1... (1-7)

C2Q2... (1-7)

C2Q3... (1-7)

C2Q4... (1-7)

All questions are answered by 10 respondants.

I've managed to find the std.dev for each question in SAS JMP:

Column 1:

1: 3

2: 3

3: 4

4: 5

5: 5

6: 6

7: 6

8: 7

9: 7

10: 7

My question is: How should I find the std.dev for the whole category?

Put all 4*10 answers in column 1?

Re: Combining multiple std.devs or combining different datas

Thanks again!

I think I've got a bit different version (JMP 10), but I put Score in Y and sorted by "category, question" and then only "category". I think I got it!

Thanks for the help.

Re: Combining multiple std.devs or combining different datas

The standard deviation of all responses in category 1 was that value. (In your example, there would be 10*7 = 70 responses in category 1... so the first 70 rows of the data table are all going to have this same value in the "Cat Stddev" column. If you used the summary platform instead of the formula approach, you'd get the following table:

Cheers,

Brady

Re: Combining multiple std.devs or combining different datas

Hey Brady!

Thanks for a great answer!

Could you please explain what the numbers in two extra columns indicates? Why "1.501192665"?

Thanks.

Re: Combining multiple std.devs or combining different datas

One easy way to do this is to put the data into 4 columns: Category #, Question #, Respondent# and Score. (See table below.)

Then create 2 new columns and add the formulas shown below.

For the standard deviation of the category:

For the standard deviation of the question (within category):

Starting with the same table, you could alternatively use Table->Summary, finding the stddev of Score, grouping once by category, and again by category and question.

Cheers,

Brady