Hi ,
I modified a valid and reliable instrument used by other researchers by adding sections to fit in my study.
The original one yielded two factors : USE and SKILL (Researchers reported them); however, they used different scales for each factor . For Factor 1, the scale ranged from Never till several times a day (6 scales) that showed frequency the respondent used that factor . As for the second factor, a scale from 1 till 7 was used to measure skills [ (1 (beginner) 7 (Expert)).
I added part 3: where respondents have to report on their attitude towards their behavior. (students have to check their preferences). I took these items from another research study.
My question is will it be a problem if I try to group factors based on review of literature? Can I do factor analysis if I have different parts but each part has a different subscale?
Regards,
Hi Hoda Kain ;
you can use EFA for gruoping items. It can be differ from other researches.
You can factor analyze each part seperetaly but if you analyze them all together, the results can be differ. I think, you may analyze all items together. So, you can see which item load which construct.
All instruments require empirical validation.
That said, IRT ideal for new measyre validation, content-based or otherwise.
If you use correlation matrix, different number of levels, different range of variables etc. will not a problem in factor analysis. But I can't understand, whether you are going to use exploratory or confirmatory FA. Using EFA for validation your factor model can lead to unappropriate results when your data is not perfecly fit to your model
Matthew Kerry, Aleksandr Raykov , and Abdullah Faruk Kılıç thanks all for your valuable comments. The instrument I am using has more than one factor, which I took from different researchers. I mean each factor is adopted from a different study and it was checked for validity and reliability by original authors. However, to fit in the scope of my study ( For example, I am measuring Internet Use (factor 1), Internet Skills (factor 2) , and attitude (factor 3). Each has been adopted from a different study. In my instrument each factor is written separately.
I read in research books (Pallant) that I have to check for EFA since it is a new instrument now.
Aleksandr Raykov , Could you clarify this point please? " Using EFA for validation your factor model can lead to unappropriate results when your data is not perfecly fit to your model "
Sir,
you can use any part of the scale, you just need to factor analyze it first. if the factors are formed in the way, you interpret correct, you can pursue with CFA finding model fit and validation.
The only thing here that i am not in agreement here is using different pointer scales, like 6 and 7. When you are already validating the scale then you can take all the statements on 6 or 7 pointer and then perform its EFA and CFA. If the factors are formed correctly and later on proves their validity are fit to use...
Hoda Kain ,
EFA is exploratory method, it conducts non-directional search (without hypotheses) and overfits to your sampling data. It is rather difficult to reproduce supposed factor structure, when factors are constructed blindly, and all slighting deviations have sufficient influence to results. If you have hypotheses about number of factors and their components, you should use confirmatory factor analysis (CFA).
- Badges
- Science topic
- Similar topics
- Instrumentation Engineering
- Instruments