Log in
Register|
Helmut ★★★   Vienna, Austria, 2008-08-21 16:10 (6095 d 06:16 ago) Posting: # 2217 Views: 8,492 |
|
|
Dear all, I was asked asked to design a pilot study to assess the CVintra of an innovator’s product only (the generic is not ready, actually it’s some kind of feasibility issue). Problems arise from the ‘pseudo randomisation’ – there’s only one product and sequences in a conventional 2×2 cross-over will be fixed a priori like this:
S/P | I IIOf course the assignment to ‘R1’ is ‘R2’ is arbitrary; any other then the chosen one in the protocol will result in another PE and CVintra… Any ideas/suggestions? — Dif-tor heh smusma 🖖🏼 Довге життя Україна! ![[image]](https://static.bebac.at/pics/Blue_and_yellow_ribbon_UA.png) Helmut Schütz ![[image]](https://static.bebac.at/img/CC by.png) The quality of responses received is directly proportional to the quality of the question asked. 🚮 Science Quotes |
|
ElMaestro ★★★ Denmark, 2008-08-21 16:32 (6095 d 05:55 ago) @ Helmut Posting: # 2218 Views: 6,623 |
|
|
❝ Any ideas/suggestions? Dear HS, could you clarify: What is R1 and R2 - from your description I don't get it. EM. |
|
Ohlbe ★★★ France, 2008-08-21 17:17 (6095 d 05:09 ago) @ Helmut Posting: # 2219 Views: 6,750 |
|
|
Dear Helmut, If I get the point correctly, your idea is to reproduce the same ANOVA model as what you would have in a conventional 2x2 study, is that right ? I mean, to introduce in the ANOVA the sequence effect (normally TR or RT) which you wouldn't have in your study as you only have RR ? If I look at slide 63 of your lecture, sequence effect can be due to either a real sequence effect (you wouldn't have one with a single RR sequence), a failure of randomisation (not with no randomisation), a true carryover effect (you know enough about BE trials to plan your trial properly, at least if the sponsor listens to you  ), or a formulation by period interaction (which I'm not sure you would be able to detect with your artifical R1R2 / R2R1 randomisation). Adding a "randomisation" in your trial and in your model would indeed introduce a "sequence" effect in your ANOVA, and therefore reduce the MSE somehow (and then reduce the intra-CV), but wouldn't it just be an artefact ? Couldn't you then be under-estimating your intra-CV ?I'm just thinking aloud, you know much more about these things than I do. Best regards Ohlbe |
|
Helmut ★★★ Vienna, Austria, 2008-08-21 19:22 (6095 d 03:05 ago) @ Ohlbe Posting: # 2222 Views: 6,696 |
|
|
Dear Ohlbe, ❝ If I get the point correctly, your idea is to reproduce the same ANOVA model as what you would have in a conventional 2x2 study, is that right ? Yes, but it's bullshit. DLabes already gave a nice explanation in his post. Thanks for thinking aloud - sometimes you don't to see the wood for the trees...  — Dif-tor heh smusma 🖖🏼 Довге життя Україна! Helmut Schütz The quality of responses received is directly proportional to the quality of the question asked. 🚮 Science Quotes |
|
d_labes ★★★ Berlin, Germany, 2008-08-21 18:59 (6095 d 03:28 ago) @ Helmut Posting: # 2220 Views: 6,849 |
|
|
Dear HS, (emphasis by me) ❝ Problems arise from the ‘pseudo randomisation’ – there’s only one product and sequences in a conventional 2×2 cross-over will be fixed a priori like this: ❝ S/P | I II❝ -----+-------❝ 1 | R1 R2❝ 2 | R2 R1❝ Of course the assignment to ‘R1’ is ‘R2’ is arbitrary; any other then the chosen one in the protocol will result in another PE and CVintra… I cannot imagine any criterion to decide what R1 and R2 is, if it is the same product, same batch and so on. What you have is only S/P | I IIwith µR the mean for R (your Reference product) and p a period effect. Within a subject you have to add an error term to specify the full model. Thus you cannot and need not randomize any arbitrary sequences. I would just run an ANOVA with a fixed period effect and a random subject effect. The residual error is then what you are looking for, I think. — Regards, Detlew |
|
Helmut ★★★ Vienna, Austria, 2008-08-21 19:16 (6095 d 03:11 ago) @ d_labes Posting: # 2221 Views: 6,685 |
|
|
Dear DLabes! ❝ I cannot imagine any criterion to decide what R1 and R2 is, if it is the same product, same batch and so on. Yes, true. I thought I may overcome these problems in ‘fixing’ treatments – of course arbitrarily – in order to stay within a conventional layout. ❝ What you have is only ❝ S/P | I II❝ ------+---------❝ RR | µR µR+p❝ with µR the mean for R (your Reference product) and p a period effect. ❝ Within a subject you have to add an error term to specify the full model. ❝ ❝ Thus you cannot and need not randomize any arbitrary sequences. ❝ ❝ I would just run an ANOVA with a fixed period effect and a random subject effect. The residual error is then what you are looking for, I think. Thanks, that helped a lot! — Dif-tor heh smusma 🖖🏼 Довге життя Україна! Helmut Schütz The quality of responses received is directly proportional to the quality of the question asked. 🚮 Science Quotes |
Ing. Helmut Schütz