Test scoring

[DVGlasgow at aol.com]

In a message dated 11/6/02 12:42:49 AM, 
use-revolution-request at lists.runrev.com writes:

<< >> I am working on an app to score a test. It is long; nearly 600 items, 
>> all True/False/Unanswered. It will be scored on a number of scales. 
>> Any suggestions for structuring this data efficiently? >>

not sure why this post was repeated, but I do the same kind of thing. As 
David and Yves contributions imply, speed is not an issue.  The approach i 
use is much more inelegant, but makes psychometric/scale development and 
modification much easier.  On a scoring card I have small scrolling fields 
named after each of the scales.  In each I have a list of the items 
contributing to the scale.  The first item on each line is the (test) item 
number.  The second is a modifier value which operates on the item response.  
So, for example, you can weight particular items, or reverse score them.  
Most importantly, you can tweak weightings during test development, and 
double check how each scale score is derived.  A score button simply sends a 
message to each field in turn which then 'scores itself up, placing weighted 
item scores in the third item of each line.  Summate these and then bingo, 
there's your scale score.  After you are absolutely sure each scale is scored 
correctly, you can replace individual scripts in the fields with a single 
background script.  To be honest though, it all happens so quick, i often 
don't bother.

For smaller questionnaires (for people with a learning disability) I place 
each item on a separate card in a background field (as opposed to a scrolling 
group of radio buttons above).  The user reponse itself triggers a script 
which adds or subtracts from the relevant scale(s) in real time.  The 
closecard meassage is used to send the result to scale accumulators

If you are converting scale scores to Z  or T scores using standardisation 
table, you can place tab delimited data in a field with scales as columns and 
raw scores in rows (except for the first which contains T or Z scores).  Two 
tips here.  

1 / Most standardisation table columns have unequal numbers of rows.  Make 
them all the same by padding out the rows after the last value with the last 
value.  This means that 'get word j of line i' will always work. 

2/  Entering a standardisation table in the first place is a real pain, and a 
very easy source of error.  Getting your machine to read the table to you 
while you scan the original is a really good way to error check.

Best wishes,

David Glasgow
Home/ forensic assessments --> <A HREF="http://members.aol.com/dvglasgow/">
DVGlasgow </A>
Courses --> <A HREF="http://www.i-Psych.co.uk">i-Psych</A>