Is it acceptable to use a table to convert scaled scores to percentile ranges? I am writing reports for medics and have scores from a range of different neuropsych tests. Some of the tests provide scaled scores and others standard scores / percentiles. I want to present all results as percentiles for ease of interpretation by medics.
Gerard Victory. I see your problem. It's not unlike what some of us had to do when having a range of markers assessing students' work and it was obvious that some markers were much harsher than others. Back then we used to convert each marker's marks to T-scores, and work from there. As the course coordinator, I kept a close track on what was happening, and it seemed to work.
I'm going to stick my neck out and say that I think what you're proposing would be OK. However, if I were you, I'd look carefully at what was going on from time to time to ensure that no serious distortions were occurring.
But I'd be interested in other people's opinions.
All the best with what you're doing.
I agree with Robert, this can serve as a guide:
Article Percentiles Please: The Case for Expressing Neuropsychologic...
Article Developments in Neuropsychological Assessment: Refining Psyc...
https://pdfs.semanticscholar.org/44d9/ecd9c226bdd98b466b14b6c10319676ab6c2.pdf
Assuming the tests that you use have these tables, then it is fine. If you just have the raw or z score though and don't have a conversion table then you it will be problematic. THey are unlikely to be normally distributed, so saying a z score of 1.96 equates with a percentile of 97.5 would be inappropriate.
Another point is that if your goal is to compare percentiles across tests it is necessary that these percentiles have been established with the same populations. If you are saying someone is in the 90th percentile of general population on test X, but in the 50th percentile of a clinical population on test Y, you can't say one is higher than the other.
Hello Gerard,
To the good answers already given, I'd simply like to make two points about percentiles:
1. A virtue of percentile ranks is that, most people can understand them (though the first exposure will require a brief explanation, so as not to confuse them with "percent correct", for example).
2. A drawback of percentile ranks is that, for any distribution other than a uniform (rectangular shape), the distances between adjacent percentile values are not uniform. For example, in a normal distribution, the difference between the 50th and 60th percentile points only covers a distance of 0.25 standard deviations, whereas the difference between 98th and 99th percentile points covers a distance of 0.27 standard deviations. So, the same sheer distance separating the points can represent 10 percentile points difference or 1 percentile point difference!
The person who is unaware of this phenomenon may over- or under-interpret specific differences between percentiles (other than they do represent ordered values).
Good luck with your work.
Dear collegues,
There are a software that convert percentile score in scaled score?
Thanks
- Badges
- Science topic
- Similar topics
- Psychology
- Psychometrics