search

xuyiqing / fect

fixed effect counterfactual estimators
Other
53 stars 20 forks source link

How to check the value of the tuning parameter lamda in "fect" #25

Closed Yanxia-Yu closed 1 year ago

Yanxia-Yu commented 2 years ago

Hi Yiqing,

I've got two questions:

1) When using the "fect" package in R to produce the MC estimates, a list of "lambda.norm" values is produced by iteration. What does "lambda.norm" mean?

2) I've found that by specifying the range of lambda, we can avoid potential over-fitting with the MC estimator. Yet I don't know how to print the value of the lambda after the estimation. Please tell me how to do that?

My data and code are attached to this question. Please have a look and rely me at your earliest convenience. Thank you.

Regards,

Yanxia Data & Code.zip

xuyiqing commented 2 years ago

Hi Ziyi, could you take a look?

On Wed, Jul 6, 2022 at 1:40 AM Amy @.***> wrote:

Hi Yiqing,

I've got two questions:

1.

When using the "fect" package in R to produce the MC estimates, a list of "lambda.norm" values is produced by iteration. What does "lambda.norm" mean? 2.

I've found that by specifying the range of lambda, we can avoid potential over-fitting with the MC estimator. Yet I don't know how to print the value of the lambda after the estimation. Please tell me how to do that?

My data and code are attached to this question. Please have a look and rely me at your earliest convenience. Thank you.

Regards,

Yanxia Data & Code.zip https://github.com/xuyiqing/fect/files/9053141/Data.Code.zip

— Reply to this email directly, view it on GitHub https://github.com/xuyiqing/fect/issues/25, or unsubscribe https://github.com/notifications/unsubscribe-auth/AB2PKGGCQ3IZ6ILZHO3BTSTVSVA7VANCNFSM52Y457MA . You are receiving this because you are subscribed to this thread.Message ID: @.***>

-- Sent from IPhone. Please pardon typos and brevity.

lzy318 commented 2 years ago

You can use the parameter lambda.cv, which is the parameter used in the paper.

lzy318 commented 2 years ago

The results of MC cross-validation are saved in out$CV.out.mc.

Yanxia-Yu commented 2 years ago

You can use the parameter lambda.cv, which is the parameter used in the paper.

Thanks a lot. I now understand how to print the optimal lambda value (out$lambda.cv). Yet I'm still confused about the "lambda.norm" values in the output . MSPE is used as a criterion to produce the optimal "lambda.norm". My question is: what is "lambda.norm"?

lzy318 commented 2 years ago

Lambda.norm is the lamba.cv divided by the maximum eigenvalue of the outcome matrix. It is just one way of normalization.

Yanxia-Yu commented 2 years ago

Lambda.norm is the lamba.cv divided by the maximum eigenvalue of the outcome matrix. It is just one way of normalization.

Thank you. ; )