tm
New Member
Posts: 1
|
Post by tm on Feb 20, 2020 20:47:29 GMT
Hi,
Using the built-in calculations, I was creating a variable per scale that calculated the composite score. However, I discovered in testing/piloting that blank answers were included in the score as the maximum possible value rather than as blanks. For example, the composite score of the selection below gets calculated as (1 + 1 + 5 + 1 + 1)/5 = 1.8 rather than (1 + 1 + 1 + 1)/4 = 1 as I would expect. (In case that's too small to see, the scenario is: there are 5 items in a 5-point Likert scale using default score values. Items 1,2,4, and 5 are answered, but item 3 is blank.) Does anyone know if there's a way to count missing values as blanks?
I tested a number of different codes, including: l: meanVal t: set - mean $item.1 $item.2 $item.3 $item.4 $item.5
l: meanVal t: set - mean $item*
|
|
|
Post by PsyToolkit on Feb 23, 2020 12:02:08 GMT
This is a good question. I answered you already in email, but here are some general points: 1) This will only happen when you use the option "free" in scale questions. That will allow people to skip items in a scale question. For psychometrics, this is actually really problematic, because these scales are typically assuming people answer all questions. 2) In the new version of PsyToolkit, you will now get NA in your data sheet. For this use version 2.6.2 or higher. As of February 2020, this is new and can be logged in via this link (otherwise, this is a stable test version): www.psytoolkit.org/cgi-bin/psy2.6.2I would recommend not to analyze people with NA missing values. Some American IRBs insist on having this option, but I would recommend to using "free" for scale questions. If people do not want to answer key questions of your survey, just give them the option to drop out from your survey altogether. Important note: Currently, if you use mean with NA values, you get meaningless numerical results. You can prevent this by either not allowing "free" or by just making sure in your data analysis to manually remove participants with NA scores.
|
|