your results indicate that there are four factors in the data. They wont load onto one.

I dont want them to load on one factor but at least they should load on their respective factors. it doesn't make a proper pattern matrix. when i am trying to run CFA in AMOS using this data its giving me different loading. do you have any idea how to fix it ?

I do not know what exactly you want to do. however you may follow this..

If you are exploring, then you should do exploratory factor analysis. Your data may or may not behave in the same way as you have conceptualized.

If the data isn't behaving in a certain way that simply means the way you have conceptualized the construct, is different from the conceptualization of your respondents.

If all the four factors are measuring similar things, you may also see the factor correlations. if the correlations aren't strong that means your respondents have understood the concept in a different way.

If the correlations are high, you may test second order construct.

Hey Shagufta!

A) what i have understood- if i am right you expected your items to load under certain "respective factors". These factors must be based on some theory or your past literature review. Now your difficulty is that the items are not loading on the factors you wanted to.

Suggestions - You should have opted for Exploratory Factor analysis on this data , as mentioned above by usama it would help you understand how your sample has conceptualised the construct you have tried to measure , it could be different than what you believed originally to get as a result and thats fine!

But now what i can see is you have done "Principle component analysis" which is different from EFA . (Please read some articles for more clarity).

Now what can be done :- 1) try compressing the data in PCA to the number of factors you believed should have come on the basis of your underlying theory. then lets see what new pattern matrix is made.

2) If you actually need to regenerate components as done before, just be transparent with the issue. If your findings contradict previous studies, then mention it.Just do not stress on getting exactly the same factor loading as you expected. research is all about this exploring and finding something new!

B) Now this is my confusion :- You said you ran CFA on the same data and it gave you different results. Now for doing CFA - that is confirming the results of your pca you should collect new data and do on that not on same data!

c) Also please see your items as well probably your item construction phase itself needs some reconsideration as mentioned by fellow research probably there is discrepancy in the way you understood the construct and others are understanding.

Hope it helps!

Hi Shugufta,

Are you studying the validity of the measurement instrument that you use? If so, then it is theoretically preferrable to use the exploratory factor analysis. Assuming this is your goal and given that principal component analysis (PCA) and factor analysis (FA) often give similar results, I'll refer to the factors instead of components below.

It would be useful to see the full exploratory factor analysis results, because it is difficult to see what is actually happening in your data (for example, the pattern matrix is set to show only one loading for item, but what are the secondary loadings of each item on the other factors). Also knowing nothing about the items makes it difficult to see how the items could be theoretically connected and what their distribution is (continuous/ordinal/categorical). Did you study how many factors your data actually supports e.g. with parallel analysis or the scree plot?

The fundamental difference with the exploratory and confirmatory factor analyses is that they are based on different assumptions of the model that is thought to have produced the data. In exploratory FA each item has a loading on all factors, so items share a part of their variance among all factors. In confirmatory FA, factors are usually constrained to load / share variance with just one factor. Now, if the data generating mechanism is such that each item loads on only those single factors that the theory predicts, then the exploratory and confirmatory analyses should agree yielding similar results. If the theoretical model was not correct, then the two models can produce differences in the loadings. You can usually see this in the secondary loadings: if an item has large enough loadings on other factors (except those that you showed in the pattern matrix), then the confirmatory model would give different loadings, because the data is not modelled correctly.

Thus, you should not try to force the exploratory model, but rather see it as telling you that the confirmatory model may be overly restrictive to be used with your data: the theory may not be predicting accurately what happens in your target group. If your goal is to examine the structure of the measurement tool, then I would suggest you study the exploratory FA model in more detail to see where the model predictions are not met (what items cluster in unexpected factors); can you think of a reason why they do so? In this way you may be able to refine the theory underlying the measurement instrument for your specific target group.

Also, make sure that the data is ok for FA. Certain things, like negatively worded items, may result in uninterpretable model results.

Regads,

Timo

thank you all for your suggestions. i ran the same analysis using different data and it worked. i think my data has some issue that`s why the pattern matrix was not loading properly.