Log in
RegisterAnova model for batch effect [General Statistics]
Hello arl stat,
Yes, that is no issue as long as subjects are numbered uniquely.
OK, Subject(Sequence*batch) again...
You have two sequences. You have two batches. You have 25 subjects in each batch and these are uniquely coded. 1 to 25 are the subjects in batch 1 and 26 to 50 or something like that are the subjects in batch 2.
When you ask for a model with Subject(Sequence*batch) then you actually ask the engine to derive effect estimates for Subject 1 in sequence 1 in batch 1, Subject 1 in sequence 1 in batch 2, subject 2 in sequence 1 in batch 1 ... subject 50 in sequence 2 in batch 2.
That is all in all 100 effects. Add to that the verious other effects of terms like period and treatment. You realise you now are asking for a number of effect estimates that exceed the number of observations, so is a silly mess. For example, subject 50 is in batch 2 so it doesnt make sense to ask for an effect of subject 50 in sequence 1 in batch 1 and so forth.
The whole Subject(Sequence*batch) only makes sense if you code subjects nonuniquely. Then you may have subjects 1..12 in sequence 1 of batch 1, subject 1..13 in sequence 2 of batch 1, subjects 1..13 in sequence 1 of batch 2, and subject 1..12 in sequence 2 of batch 2. Suddenly it now makes sense to speak of Subject(Sequence*batch).
But of course in real life noone will number subjects nonuniquely for at least two reasons.
Let me add, it is not leading to erronoeus results per se to use Subject(Sequence * batch) when uniquely coded subjects are used because the model matrices will be reduced to full rank internally by the lexer or engine before the fit is done.
❝ In particular, the model should reflect the fact that the periods for the first group are different from the periods for the second group.
Yes, that is no issue as long as subjects are numbered uniquely.
❝ Sorry but, I did not get your explanation.
OK, Subject(Sequence*batch) again...
You have two sequences. You have two batches. You have 25 subjects in each batch and these are uniquely coded. 1 to 25 are the subjects in batch 1 and 26 to 50 or something like that are the subjects in batch 2.
When you ask for a model with Subject(Sequence*batch) then you actually ask the engine to derive effect estimates for Subject 1 in sequence 1 in batch 1, Subject 1 in sequence 1 in batch 2, subject 2 in sequence 1 in batch 1 ... subject 50 in sequence 2 in batch 2.
That is all in all 100 effects. Add to that the verious other effects of terms like period and treatment. You realise you now are asking for a number of effect estimates that exceed the number of observations, so is a silly mess. For example, subject 50 is in batch 2 so it doesnt make sense to ask for an effect of subject 50 in sequence 1 in batch 1 and so forth.
The whole Subject(Sequence*batch) only makes sense if you code subjects nonuniquely. Then you may have subjects 1..12 in sequence 1 of batch 1, subject 1..13 in sequence 2 of batch 1, subjects 1..13 in sequence 1 of batch 2, and subject 1..12 in sequence 2 of batch 2. Suddenly it now makes sense to speak of Subject(Sequence*batch).
But of course in real life noone will number subjects nonuniquely for at least two reasons.
- It is not intuitive to read the tables if not numbered uniquely.
- GCP inspectors sometimes ask for a procedure how unique subjects numbers are assigned to enrolled subjects and they give you a deficiency if you do not have a procedure for it.
Let me add, it is not leading to erronoeus results per se to use Subject(Sequence * batch) when uniquely coded subjects are used because the model matrices will be reduced to full rank internally by the lexer or engine before the fit is done.
—
Pass or fail!
ElMaestro
Pass or fail!
ElMaestro
Complete thread:
- Anova model for batch effect arl_stat 2013-03-25 12:32
![[Mix]](https://static.bebac.at/img/mix.png "open in Mix view")
- Anova model for batch effect ElMaestro 2013-03-25 15:23
- Anova model for batch effect Ohlbe 2013-03-25 16:50
- Anova model for batch effect arl_stat 2013-03-25 18:27
- Anova model for batch effect ElMaestro 2013-03-25 19:34
- Anova model for batch effect arl_stat 2013-03-26 05:35
- Anova model for batch effect ElMaestro 2013-03-26 09:04
- Anova model for batch effect arl_stat 2013-04-01 13:09
- Anova model for batch effectElMaestro 2013-04-01 13:27
- 50 x 2 x 2 = 100?? ElMaestro 2013-04-02 19:28
- Anova model for batch effectElMaestro 2013-04-01 13:27
- Anova model for batch effect arl_stat 2013-04-01 13:09
- Anova model for batch effect ElMaestro 2013-03-26 09:04
- Anova model for batch effect arl_stat 2013-03-26 05:35
- Anova model for batch effect ElMaestro 2013-03-25 19:34
- Anova model for batch effect arl_stat 2013-03-25 18:27
Ing. Helmut Schütz