I have a self-efficacy questionnaire and I want to know how can I classify the score of my questionnaire. Do I use the medium to classify the subjects into "self-efficacy" and "no-self efficacy"? Or what can I do?
With group SE scores there tends to be a bit of a ceiling effect. Our group means tend to be almost 7 on a ten point scale. We have also done a little work on seeing what confidence levels actually predict behavior change. We have chosen 7 and above but you could chose 6 and above if you wished. I would not go with any mean or median but rather with a cut off score.
Hi Alejandra,
Yes, splitting the groups at the median point is one way of doing it. However, I have been warned against doing this because when you split continuous level data (although your self-efficacy scale may not strictly fall into that category) into yes/no or high/low categories, you are losing statistical power; MacCallum et al, 2002 mention this here: -
http://www.psychology.sunysb.edu/attachment/measures/content/maccallum_on_dichotomizing.pdf
And there is a brief article on it here: - http://www.ncbi.nlm.nih.gov/pmc/articles/PMC1458573/
I think it also depends on what you are doing with it, and whether self-efficacy is a predictor or outcome variable. You could also have a look to see what how other researchers have used the self-efficacy scale you are using (or similar ones). It is difficult to say what you 'should' do without knowing more, but, for example, if you are using self-efficacy to predict another variable or outcome I would probably suggest using regression.
I'm sure there will be some more informative answers along soon, but I hope that helps for now.
Good luck,
John
Hi,
Adding to the above, you may want to explore a Rasch model approach to this question. Performing a Rasch analysis of your instrument will allow you to empirically assess how your participants can be meaningfully divided into groups based on their levels of self-efficacy (as opposed to just splitting the sample in two equally large groups, which would be the above/below median approach that you have discussed). A recent article may assist you a bit further:
http://www.hqlo.com/content/11/1/202
Best wishes,
Tore
With group SE scores there tends to be a bit of a ceiling effect. Our group means tend to be almost 7 on a ten point scale. We have also done a little work on seeing what confidence levels actually predict behavior change. We have chosen 7 and above but you could chose 6 and above if you wished. I would not go with any mean or median but rather with a cut off score.
Thanks for all your answers they have been really helpful.
I have another question. Can I do a Rasch model in STATA? or Do I have to use WINSTEPS?
Alejandra,
I am not familiar with STATA, but WINSTEPS are commonly used when applying the Rasch Model with two facets (often persons and items). When more facets are examined (e.g., persons, items, and raters), a software called FACETS is commonly used. In addition to the required software, I strongly recommend that you recruit someone to support you with the analysis and its interpretation. I know I would have been 'lost' without it....
Best wishes,
Tore
I think you're looking for a quick and simple method to identify who scored high and low in your sample on your Self-Efficacy questionnaire?
Following on from John Hudson's comments, you could very simply transform participant raw scores into z-scores. WIth these scores you can easily identify who is high or low without turning their data into categorical variables.
A z-score of 1 (high score), 2 (very high score) or 3 (exceptionally high score) = 1 , 2 or 3 standard deviations higher than your average sample score. Conversely a z-score of -1 (low score), -2 (very low score), -3 (exceptionally low score) refers to standard deviations lower than the average sample score. A score between -1 and 1 = an average score.
If you are using SPSS, there's a box you can tick when you ask for Descriptives that does this automatically. It creates a new column in your dataset with everyone's z scores.
If you're using STATA, then here's the page. http://www.ats.ucla.edu/stat/stata/faq/standardize.htm
Hope this helps!
Here are some ways you could assign your cut-points:
1) Estimate the level of the trait via concurrent validity or expert opinion for persons along various portions of the scale. Set the cut-points based on these external definitions.
2) Construct a histogram of the sample participants’ scores and see if there may be natural cut-points that develop.
3) Are there expected proportions of inclusion within each range of the scale? If so, use those proportions as a basis for your intervals.
4) If you have other measures, are the multiple measures correlated well? Locate the persons that score high on all of the measures and base the interval for scoring high on the ranges for these persons. Do the same for those that score low on all of the measures. This will give you two of the ranges.
5) Depending on the sample, the persons involved could serve as the basis for the interpretation of the scores. Rather than defining the intervals and understanding the persons within them, do the opposite. Understand the persons along the full score range and use related information about them to then define the intervals.
- Badges
- Science topic
- Similar topics
- Mathematics
- Statistics
- Applied Statistics
- Survey Design
- Questionnaire