Simulation

[seyifemickael]

I am interesting to simulate data nature for Small area estimations. However how can I write sas/IML

[PaigeMiller]

Please describe the problem in a lot more detail.

Also, why does this have to be done in PROC IML?

[PeterClemmensen]

@seyifemickael Welcome to the SAS Community 🙂

You have to describe your problem in much more detail than this. What is the problem? What have you tried already? 

[Rick_SAS · Accepted solution]

First, I suggest you read the paper "Small Area Estimation for Survey Data Analysis Using SAS Software"
(Mukhopadhyay and McDowell, 2011) which uses mixed models for SAE.

After you settle on an explicit statistical model, which includes assumptions about the spatial covariance among the small areas, then you can start researching how to simulate from that model by using standard techniques such as are described in Chapter 12 of Simulating Data with SAS (Wicklin, 2013).

This topic is large and complicated, so you should discuss it with your advisor, mentor, or manager.

[seyifemickael]

Thank you in advance for your cooperation. I have read the paper "Small
Area Estimation for Survey Data Analysis Using SAS Software" (Mukhopadhyay
and McDowell, 2011) which uses mixed models for SAE and again an idea. The
paper did the real data case for both unit level and area level SAE.
However, my interest is how to simulate those covariates as well as the
data.

[Rick_SAS]

To simulate the data, you need to specify the data generating mechanism for the spatial data. You also need an estimate of the spatial covariance matrix, which you can get from analyzing the real data that you are trying to simulate.

For a Gaussian random field, you can use the SIM2D procedure, which supports the SIMULATE statement. The documentation includes several examples.

In Chapter 14 of Simulating Data with SAS, discusses several spatial models, including

1. Simulate data from a Gaussian random field (unconditional or conditional)

2. Simulate data fro a spatial point process