topic Re: Simulation using SAS in SAS/IML Software and Matrix Computations

Simulation using SAS

GehadElsayed123 — Sun, 24 Jun 2018 00:38:59 GMT

i want to simulate , the in-control performance of the Phase I T2 chart is evaluated when the missing values are estimated by each of
the four handling methods; MSS, EM, RG and SRG. We consider the historical sample m (m=20), of size
n=1, five values for the percent of missing value k% (k% = 1%,), three values for the number of quality
characteristics p (2, 3 and 5) and four values for the correlation coefficient r(r = 0, ,). Similar conclusions from other
simulations (not shown here) are obtained when different values of p and r are considered. The value of h that produces an overall
probability of a false alarm a = 0.05, when full data is assumed, was computed using 20,000 simulation runs. Table I presents these
values of h according to the number of variables p and the number of samples m. Notice that these control limits do not depend on
the variance–covariance structure since they are calculated based on full data sets.

i do this code but i don't complete it , there are some errors

so i do not know how i solve it .

Re: Simulation using SAS

pau13rown — Sun, 24 Jun 2018 01:17:27 GMT

can you be more specific about the problem you're having. Have you read Rick Wicklin?

Re: Simulation using SAS

PaigeMiller — Sun, 24 Jun 2018 01:19:59 GMT

there are some errors

Show us the SASLOG with the errors. Click on the {i} icon and paste the relevant parts of the SASLOG into there.

Simulation using SAS

GehadElsayed123 — Sat, 30 Jun 2018 07:09:41 GMT

i do this code but i don't complete it , there are some errors

so i do not know how i solve it .

Re: Simulation using SAS

FreelanceReinh — Sat, 30 Jun 2018 11:31:33 GMT

@GehadElsayed123:

Why don't you continue the thread that you started six days ago with the identical text and (up to blank lines) identical attachment in the appropriate forum for SAS/IML? Respected experts asked you questions there and are awaiting your answers.

I'm sorry, I can't help you with this, although I like doing simulations with SAS. But I haven't been using SAS/IML for a long time and your request is unclear. For example, your (copied-and-pasted?) text mentions "five values for the percent of missing value", but specifies only one ("1%"). Same with the "four values for the correlation coefficient". The reference to "Table I" is useless as there is no such table shown in your post, nor in the attachment.

Re: Simulation using SAS

Rick_SAS — Mon, 02 Jul 2018 17:11:20 GMT

Many times the key to a successful simulation study is knowing how to generate one simulated sample before you try to simulate 10,000 samples. In your case, before you loop over a grid of parameters, know how to do each simulation separately. For example

Handling method=MSS (?what is this?)

Percent of missing value = 1%

quality characteristics p=3

and correlation coefficient r = 0.5

Only after you can successfully generate 1 sample for these conditions, should you attempt 20,000.

If you get stuck, post the code along with a verbal explanation of what the "handling method" is and what you mean by "quality characteristics".

Re: Simulation using SAS

GehadElsayed123 — Mon, 02 Jul 2018 19:32:08 GMT

i want to simulate , the in-control performance of the Phase I T2 chart is evaluated when the missing values are estimated by each of
the four handling methods; MS=mean substitution , EM=the EM algorithm , RG =regression imputation and SRG=stochastic regression imputation . We consider the historical sample m (m=20), of size
n=1, one value for the percent of missing value k% (k% = 1%,), one value for the number of quality
characteristics p (2, ) and one value for the correlation coefficient r(r = 0, ,).
The value of h= ucl that produces an overall probability of a false alarm a = 0.05, when full data is assumed, was computed using 20,000 simulation runs. Table I presents these
values of h according to the number of variables p and the number of samples m. Notice that these control limits do not depend on
the variance–covariance structure since they are calculated based on full data sets.

i do this code but i don't complete it , there are some errors , i want to know ,how to solve theses errors ?

so i do not know how i solve it .

i do this code but i don't complete it , there are some errors

so i do not know how i solve it .