I want to know if you can "compare" results of different surveys that assess the same thing (Example: quality of life) having different ranges, into one measurement:
Example:
Survey 1 --> range = 0 - 15 , result = 8
Survey 2 --> range = 5 - 25, result = 15
Survey 3 --> range = 3 - 18, result = 6
The idea is to standardized the three surveys into one scale to "compare" them.
Hello there,
I think you could try to compare effect sizes (e.g. regression coefficients). I think this is a procedure for numerous articles on the same construct (e.g. loneliness) and effect size constitutes a common denominator used for comparison purposes.
Regards, Witold
You mention the word "standardize" and one way to compare the distributions would be to convert each of them to z-scores (i.e., mean = 0 and std. deviation = 1).
Also, Witold has an interesting suggestion, since the original response scoring will not matter for linear relationships such as correlations or regressions. For example, if you have age as a variable in each data set, you could compare the relationship between age and quality of life across the three surveys.
Thank you both!
To be more specific i'm conducting a literature review . The selected studies report for each sample a mean score of satisfaction of the used instrument. Generally, these instrument have different scales for minimum and maximum so in order to create a sort of “comparison“ between all of the studies i`m looking for a method that allows me to rank them. For example:
Study n°1, instrument: A , mean score 9 (range = 0 - 12)
Study n°2, instrument: B , mean score 3 (range = 0 - 4)
Study n°3, instrument: C , mean score 20 (range = 5 - 25)
Study n°4, instrument: D , mean score 78 (range = 0 - 100)
Thanks for all of your comments!
Hello,
Study n°1, instrument: A , mean score 9 (range = 0 - 12)
9 : 12 = 0.75
Study n°2, instrument: B , mean score 3 (range = 0 - 4)
3 : 4 = 0.75
Study n°3, instrument: C , mean score 20 (range = 5 - 25)
20 : 25 = 0.80
tudy n°4, instrument: D , mean score 78 (range = 0 - 100)
78 : 100 = 0.78.
My conclusion is if there is no information on effect sizes then use this, figures seem to be corresponding with each other.
Regards, Witold
That was my first approach but i think that formula could not work in some cases like:
Study A, instrument n° 1, range = 0 - 14, mean = 7 --> 0,5
Study B, instrument n° 2, range = 5 - 14, mean = 7 --> 0,5
The range of the instruments are different, moreover, they have the same mean score and also they share the same maximum score but the score for study B should be lower than A because is closer to the minimum.
Regards!
- Badges
- Science method
- Similar topics
- Mathematical Sciences
- Statistics
- Applied Statistics
- Survey Design