Potential bias in stratified analyses which had imbalance sample ratio?

[haoduonge]

Hi all,

I would like to exam effect of two kinds of treatment (A, B) on health outcome (death vs non), stratified by a third variable (level 1, level 2, level 3).

The problem is that % by level in treatment A is so different from that in B.

For example: 89% level 1, 5% level 2, and 6% level 3 in treatment A, while 35% level 1, 50% level 2, and 15% level 3 in treatment B.

If I conduct the analyses among level 1 group, meaning 89% of the treatment A and only 35% of the treatment B. Do you think this may cause any bias? if so, can you please share some references about it?

Thanks!

[sbxkoenk]

Hello @haoduonge ,

Is there a reason why you post this question in the 'Text Analytics' board?

If not, I will move this topic to the 'Statistical Procedures' board as you will probably solve this with PROC FREQ or PROC MIXED / GENMOD / NPAR1WAY (in short, a SAS/STAT procedure), correct?

Let me know ...

Koen 

[haoduonge]

Hi Koen,

No reason, I thought this is more about the concept, not the statistical procedures.

Thanks!

[sbxkoenk]

> I thought this is more about the concept, not the statistical procedures.

That's absolutely true, but this 'SAS Text and Content Analytics' board is about text parsing and natural language processing (NLP) and text mining [i.e. artificial intelligence technology to transform the free (unstructured) text in documents and databases into normalized, structured data suitable for analysis].

I will move it as "regular statisticians" do not frequently look in here.

Koen

[sbxkoenk]

Moved by me to this 'Statistical procedures' board (under Analytics header).

Coming from 'SAS Text and Content Analytics' board (also under Analytics header).

Koen