statistical question
Locked
statistical question
 
|
Good morning. I have a quick question. I have a sample of individuals that have been assessed by two different instruments that measure risk of absconding. The risk of absconding is measured in decile format (1-10) and each individual is given a score. We also know the sample’s absconding rate. I ran crosstabs of risk of absconding by the actual absconding rate for each instrument and plotted the values on a chart by deciles. The lines for both instruments do not cross. People find that curious and say that because the same population is being studied, they should cross. However, I looked at how the population was distributed by decile by both instruments and they are distributed different. In one instrument the population is weighted to the lower deciles while in the other the population is weighted to the higher deciles. In other words, the same person can be assessed as a 1 in one instrument and a 6 in the other. Am I correct in assuming that the lines may not cross given the distribution of the instruments? Thanks |
Administrator
|
"People find that curious and say that because the same population is being studied, they should cross. "
WHY should they??? Sounds like the 2 instruments are positively correlated (which they should be). What does a raw scatterplot look like prior to throwing away *much* of your information by collapsing to deciles. What is the point of binning the data anyways? It is usually a TERRIBLE idea to do so! ----
Please reply to the list and not to my personal email.
Those desiring my consulting or training services please feel free to email me. --- "Nolite dare sanctum canibus neque mittatis margaritas vestras ante porcos ne forte conculcent eas pedibus suis." Cum es damnatorum possederunt porcos iens ut salire off sanguinum cliff in abyssum?" |
|
|
In reply to this post by FernandezLanier, Adriana (DCJS)-2
"You keep using that word. I think it does not mean what you think it means."
- from "The Princess Bride." When you state that the risk is "measured in decile format", you are asserting that you have collapsed some raw scores into categories that each have 10% of the total N. Lines constructed from scales treated that way *would* cross. But you go on to say that you "looked at how the population was distributed by decile by both instruments and they are distributed different." That pretty-much rules out the definition of having 10% in each, which is what it takes to make them "deciles." What you have are two scales that run from 1 to 10, with different definitions or anchors, or whatever. Stop using the word "deciles" and you will be okay. -- Rich Ulrich Date: Tue, 12 Jun 2012 08:50:28 -0400 From: [hidden email] Subject: statistical question To: [hidden email] Good morning. I have a quick question. I have a sample of individuals that have been assessed by two different instruments that measure risk of absconding. The risk of absconding is measured in decile format (1-10) and each individual is given a score. We also know the sample’s absconding rate. I ran crosstabs of risk of absconding by the actual absconding rate for each instrument and plotted the values on a chart by deciles. The lines for both instruments do not cross. People find that curious and say that because the same population is being studied, they should cross. However, I looked at how the population was distributed by decile by both instruments and they are distributed different. In one instrument the population is weighted to the lower deciles while in the other the population is weighted to the higher deciles. In other words, the same person can be assessed as a 1 in one instrument and a 6 in the other. Am I correct in assuming that the lines may not cross given the distribution of the instruments? Thanks
|
| Free forum by Nabble | Edit this page |