I am conducting a CFA through AMOS, my factor ladings are low or in negative range. Does this mean the data is having problem or can I approach the CFA differently? should i recode negative statements? i do not want to remove the statements.
My scale is based on 67 items all items chosen from literature or Alpha value is also good. sample size is 405.
@Amna Below thread on Research Gate could possibly help your cause👇🏼
https://www.researchgate.net/post/What-does-a-negative-value-for-factor-loading-mean
Depends on the type of study you want to carry out. Having stated clearly that you don't want to leave any statement out, then continue with your work. The low or negative loading just shows the items do not explain the factor well enough.
Look at the text by Kaplan: https://methods.sagepub.com/book/structural-equation-modeling
Hope this helps...
Hello Amna,
I'd need more information in order to give you an unambiguous recommendation.
What I can say is, if your scale includes both positive and negative valence/polarity variables (or items), and you have not recoded the variables to reflect a common valence, then this can be an issue with respect to obtaining easily understood results. The closer the ratio of positive valence to negative valence variables/items is to 1:1, the more challenging the results can be.
All this assumes, of course, that your data set represents valid responses (and not random or nonsensical response patterns) from your participants, that you've coded each of the variables appropriately, and set up your CFA model appropriately.
Good luck with your work.
It means that negatively loaded items measure the opposite pole of your intended measured construct. Thus you have to subtract this negative item loading to the total. For example, you developed a measure for self-esteem. You have ten-item, after the results of factor analysis you found that your scale is unidimensional and one item has a negative loading and its value is -.79. Then ıf you want to calculate the total score of a participant you have to sum up the first nine-item and subtract this negatively loaded item. This will give you the true score of a participant.
From Distefone and colllegue words:
If an item yields a negative factor loading, the raw score of the item is subtracted rather than added in the computations because the item is negatively related to the factor. For this method (as well as for the following non-refined methods) average scores could be computed to retain the scale metric, which may allow for easier interpretation. Also, average scores may be useful to foster comparisons across factors when there are differing numbers of items per factor. You can think that similar to reverse-coded items.
I found this example useful to this question, as practically experienced this issue.
David Morse thanks for your response. Yes, scale includes both positive and negative statements as i am using big five personality traits inventory or information privacy concerns related questions. I did not re-code the data faced same issue in EFA too.
Now what should i do to make the model loading better. I am using all these first time.
@Amna Farzand As your data consists of negative and positive polarity variables, you need to reverse code negative polarity variables (e.g. 1>5, 2>4, 3>3, 4>2, and 5>1 if your data is on five-point Likert scale).
Hello again Amna,
First, I'd suggest reverse coding the negative polarity items (as suggested by Imran Anwar).
However, no matter what you do, there is always the possibility that your sample is sufficiently distinct and/or there were cultural/language differences such that the nature of the relationships among individual items/questions/stimuli were different than what has been observed among other samples.
(I'm presuming here that your are using a previously developed measure, and not originating your own.)
If your goal is a version that works satisfactorily to represent the "Big Five" framework, for your sample (and, presumably, the population it represents), then yes, it is possible that you may have to rethink the assortment of items. Some might need to be discarded, reworded, or replaced outright.
Good luck with your work.
- Badges
- Science topic
- Similar topics
- Engineering
- Electrical Engineering