Hi everyone, I am searching for some toy models in R. My goal is do to model checking. For example, My data come from statistical model N(5, 2), with n=100, call this model_1 Then, I add bias to that data with N(3, 1), with n=100, call this model_2 Ultimately, I want to see model_1+ model_2 gives good prediction or perhaps parameter estimation. I think this is a pretty standard statistical analysis problem? How do people on this list deal with it? Any suggestions? Thank you, Mike
What are some toy models I can use in R?
3 messages · S Ellison, C W
-----Original Message----- My data come from statistical model N(5, 2), with n=100, call this model_1 Then, I add bias to that data with N(3, 1), with n=100, call this model_2
Do you mean you have data from N(5,2) that has had data from N(3,1) added to it, or that you have two different sets of data? Or do you mean that you want to know how to generate such data?
Ultimately, I want to see model_1+ model_2 gives good prediction
If you generate random data correctly following a model, the model will indeed predict the data pretty well. But under those circumstances it seems redundant to ask the question. Were you thinking of fitting a (possibly different) model to the data at some point ? If so, what model would you want to fit? And what would you want to predict from it?
or perhaps parameter estimation.
What parameters do you want to estimate?
I think this is a pretty standard statistical analysis problem?
Unclear on that at present. See above.
*******************************************************************
This email and any attachments are confidential. Any use...{{dropped:8}}
On Wed, May 25, 2016 at 1:13 PM, S Ellison <S.Ellison at lgcgroup.com> wrote:
-----Original Message----- My data come from statistical model N(5, 2), with n=100, call this
model_1
Then, I add bias to that data with N(3, 1), with n=100, call this model_2
Do you mean you have data from N(5,2) that has had data from N(3,1) added to it, or that you have two different sets of data? Or do you mean that you want to know how to generate such data?
I generate toy data from N(5,2). X input will be the same for model_1 and model_2, say, seq(-3, 3, by=0.01).
Ultimately, I want to see model_1+ model_2 gives good prediction
If you generate random data correctly following a model, the model will indeed predict the data pretty well. But under those circumstances it seems redundant to ask the question. Were you thinking of fitting a (possibly different) model to the data at some point ? If so, what model would you want to fit? And what would you want to predict from it?
I only use model_1, model_2 for data generation. I am using machine learning methods for estimation and prediction to see if the method is good and robust. Then, check performance by comparing results with the known model_1 and model_2.
or perhaps parameter estimation.
What parameters do you want to estimate?
I think this is a pretty standard statistical analysis problem?
Unclear on that at present. See above.
*******************************************************************
This email and any attachments are confidential. Any u...{{dropped:13}}