Prev: Using PROC HTTP from mainframe SAS
Next: Array use?
From: lyw8 on 14 Dec 2009 16:39
A chemist needs to estimate sample size to compare three calibration
curves of 11 standard concentrations for chemical analysis for three
different matrices. The null hypothesis is that there is no difference
in the intercept and slope of the three calibration linear regression
curves. The chemist would like to know the number of repeated
experimental run sequences needed in order to capture the possible
between run variation.

=20

Response will be the dependent variable, the experiment will look like
as followed with two matrices and two run sequences only: =20

=20

Matrix

Run sequence

Concentration=20

Response=20

1

1

1

=09
1

1

2

=09
1

1

3

=09
1

1

4

=09
1

1

5

=09
1

1

6

=09
1

1

7

=09
1

1

8

=09
1

1

9

=09
1

1

10

=09
1

1

11

=09
1

2

1

=09
1

2

2

=09
1

2

3

=09
1

2

4

=09
1

2

5

=09
1

2

6

=09
1

2

7

=09
1

2

8

=09
1

2

9

=09
1

2

10

=09
1

2

11

=09
2

1

1

=09
2

1

2

=09
2

1

3

=09
2

1

4

=09
2

1

5

=09
2

1

6

=09
2

1

7

=09
2

1

8

=09
2

1

9

=09
2

1

10

=09
2

1

11

=09
2

2

1

=09
2

2

2

=09
2

2

3

=09
2

2

4

=09
2

2

5

=09
2

2

6

=09
2

2

7

=09
2

2

8

=09
2

2

9

=09
2

2

10

=09
2

2

11

=09

Will a multiple regressin including matrix ( 3 levels), conc
(continuous) , and their interaction between matrix abnd concentration
be the right mdoel ? The problem is to determine the sampe size of the
run sequency and whether porc power with multreg option will be the
right one to use. If so, what parameters need to be specified ? =20

Thank you.=20

Lee-Yang Wong
 | 
Pages: 1
Prev: Using PROC HTTP from mainframe SAS
Next: Array use?