0
$\begingroup$

I have a dataset with 4712 records. Label Yes - 1558 records and Label No - 3554 records.

I read online that 1:10 rule is based on the frequency of lower occurring class.

In my case, frequency of lower occurring class is 1558

According to 1:10 rule, am I right to understand that it is calculated like 1558/10 = 155.8 further equals 150 predictors

So In my logistic regression, I can use 150 variables/input features to the model without the risk of overfitting. Am I right?

By any chance do we also have to look at the frequency of the other (high occuring) class to determine the no of predictors that I can use? If yes, can you share me as to what has to be done to determine the predictor count?

I am aware that we could also use 1:20 or 1:50 rule. But my question is mainly on

1) Whether is there any other consideration for determining the number of predictors in logistic regression model?

2) How do people calculate min sample size required based on this?

Can someone help me with this?

$\endgroup$
  • 1
    $\begingroup$ What is "EPV" ? $\endgroup$ – Leevo Jan 2 at 16:39
  • $\begingroup$ What do you mean with "min sample size"? Could you elaborate on that? $\endgroup$ – Leevo Jan 2 at 16:41
  • $\begingroup$ Hi, can you share the link where you read this. $\endgroup$ – cap Jan 2 at 20:11
  • $\begingroup$ en.m.wikipedia.org/wiki/One_in_ten_rule Thisvwhere I read this $\endgroup$ – The Great Jan 2 at 21:00
  • $\begingroup$ By sample size, I mean the number of records which is sufficiently for analysis. I mean we can't have 20 records and 60 predictors. Right? It will definitely overfit. So I guess there is something called data hungry models which can help us know this. $\endgroup$ – The Great Jan 2 at 21:24
1
$\begingroup$

1) Whether is there any other consideration for determining the number of predictors in logistic regression model?

The right number of predictors depends on your data and on your theories about data, and on that only. All these rules of thumb seem completely arbitrary to me, and they lack any scientific ground.


2) How do people calculate min sample size required based on this?

What do you mean with "min sample size"?

EDIT: Of course, the number of explanatory variables cannot be larger that the number of observations. Based on this answer, a rulo of thumb is to use at least 10-20 observations for each variable. I wish also to stress one point: they must be useful variables, i.e. variables with actual explanatory power. If a variable is a linear combination of others, it won't improve the model by any means, and statistical softwares such as R would delete one automatically.

However, my suggestion is to try to always employ all the data available, don't just stick to a minimum threshold.

| improve this answer | |
$\endgroup$
  • $\begingroup$ This is where I read this.. en.m.wikipedia.org/wiki/One_in_ten_rule. By sample size, I mean how many records can we have for analysis? Meaning if I have only 20 records, it doesnt make sense to have 60 covariates. Right? So is there a way to calculate this sample size? Isn't called data hungry models $\endgroup$ – The Great Jan 2 at 20:59
  • 1
    $\begingroup$ Upvoted For the help $\endgroup$ – The Great Jan 2 at 21:00

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.