Re: Treatment for acute anterior cruciate ligament tear: five year outcome of randomised trial
A brief rejoinder to Jonas Ranstam,
Is this really so difficult to understand? Well, apparently it is, so let me summarize in order to make it completely transparent. Validity and Rasch models are two different, but very strongly related issues. One of the purposes of a careful item analysis by Rasch models is to provide evidence of both validity and sufficiency. If the Rasch model is rejected, the evidence against the model will tell us whether we have validity issues (e.g., multidimensionality, local dependence, or DIF) or whether the problem is just a question of the wrong choice of probability functions. If this explanation is still not clear, I suggest that you consult one of the many books on Rasch and IRT models. It often helps to know what we are talking about.
Furthermore, at the end of the day, it is the obligation of those who use an instrument to provide solid evidence that measurement is not confounded and therefore that the instrument does not systematically understate or exaggerate treatment effects. The rest of the scientific community may help by pointing out that something appears to be amiss, even though such help is rarely appreciated. Please recall, that this is research and not law. In law, you are innocent until proven guilty. In research, you are guilty, or at least under suspicion, until you yourself have proven that you are innocent. The validity of KOOS has been challenged. It is up to you, and nobody else, to provide evidence supporting your claims on behalf of KOOS, since it is you who use KOOS. You can do this in two ways. By referring to papers where KOOS has been found to be valid in a population that is similar to your study population or by publishing the results of your own item analysis by either IRT models or by confirmatory factor analysis. If you cannot provide positive evidence of validity, your claims on behalf of treatment effects on KOOS are open to suspicion.
Rapid Response:
Re: Treatment for acute anterior cruciate ligament tear: five year outcome of randomised trial
A brief rejoinder to Jonas Ranstam,
Is this really so difficult to understand? Well, apparently it is, so let me summarize in order to make it completely transparent. Validity and Rasch models are two different, but very strongly related issues. One of the purposes of a careful item analysis by Rasch models is to provide evidence of both validity and sufficiency. If the Rasch model is rejected, the evidence against the model will tell us whether we have validity issues (e.g., multidimensionality, local dependence, or DIF) or whether the problem is just a question of the wrong choice of probability functions. If this explanation is still not clear, I suggest that you consult one of the many books on Rasch and IRT models. It often helps to know what we are talking about.
Furthermore, at the end of the day, it is the obligation of those who use an instrument to provide solid evidence that measurement is not confounded and therefore that the instrument does not systematically understate or exaggerate treatment effects. The rest of the scientific community may help by pointing out that something appears to be amiss, even though such help is rarely appreciated. Please recall, that this is research and not law. In law, you are innocent until proven guilty. In research, you are guilty, or at least under suspicion, until you yourself have proven that you are innocent. The validity of KOOS has been challenged. It is up to you, and nobody else, to provide evidence supporting your claims on behalf of KOOS, since it is you who use KOOS. You can do this in two ways. By referring to papers where KOOS has been found to be valid in a population that is similar to your study population or by publishing the results of your own item analysis by either IRT models or by confirmatory factor analysis. If you cannot provide positive evidence of validity, your claims on behalf of treatment effects on KOOS are open to suspicion.
Competing interests: No competing interests