Check the factor loading first. Factor loading should be more than .40 and above. If the factor loading is sufficient you do not need to drop out the mentioned item. You can also find useful knowledge in Andy Field's SPSS book.

You can also use the website:

https://stats.stackexchange.com/questions/300168/dropping-items-based-on-communalities-in-factor-analysis

Thank you Sevilay Kilmen

In my component matrix, that item which has low communality value, has also a factor loading of .358 whereas the rest have .780+

Child (2006) suggested removing those items which have communality value less than 0.20 in the dimension reduction technique. Low commonality value of an item represents a poor ratio of its unique variance to its shared variance. This indicates such an item could not contribute to explaining the construct adequately. You can read the article by Peter Samuels (2017) for further understanding.

References:

• Child, D. (2006). The Essentials of Factor Analysis. 3rd ed. New York: Continuum.
• Samuels, P. (2016). Advice on Exploratory Factor Analysis. Birmingham City University: In Progress. DOI:10.13140/ RG.2.1.5013.9766

Maria Antoniou, I wonder whether the following article might shed some light on your situation:

Gaskin, C. J., & Happell, B. (2014). On exploratory factor analysis: A review of recent evidence, an assessment of current practice, and recommendations for future use. International Journal of Nursing Studies, 51, 511–521. https://doi.org/10.1016/j.ijnurstu.2013.10.005

The item that is "unusual" does seem to me to be worth removing (disappointing for you, I guess) - though perhaps look at it closely to see whether it is interesting in any way, e.g., why it does not link up with the other items on its factor. Maybe, if you have an opportunity, and the item captures something important, you could create additional items to accompany it.

All the best with your research.

Thank you Sayoni S. and Robert Trevethan for your suggestions. Will have a look. However, my supervisor did say it can be removed but also I can maintain it and state that my pearsons' Rs is still high and justify not deleting the invalid variable on that? I am very torn and I would like to keep it as in my opinion is an important variable

In order to answer this question I would want to know what you plan to do with the factor scores, and whether you believe this factor has the same meaning without this item.

@Maria...look into the Average Variance Extracted (AVE) score of the construct including all the five items....if it is 0.50 or above, then I guess you can retain the item. AVE represents convergent validity of the construct...and score of 0.50 and above indicates that all the items, collectively, explain at least 50 % of variance..

Sayoni S. thank you that's very helpful. unfortunately my AVE is only 0.14

AVE score (convergent validity) is important Maria to establish the psychometric property of the construct. Address this issue adequately. Low AVE score may raise several questions on validity of the measurement scale used.

Sayoni S., I don't like to expose my ignorance publicly, but unfortunately I have to confess that I've not seen mention of AVE scores before.

Would you be kind enough to explain how / where these scores are obtained, please? Are they available from SPSS output?

Thanks.

Robert Trevethan ...Hello Robert, of course, I'd be happy to explain the concept of convergent validity and Average Variance Extracted (AVE) score. This score is based on item loadings of a factor in a construct. If you could message and give me your email id, it will be easier for me to get in touch and expand on the topic (including how to calculate AVE).

Thanks and regards

Sayoni

Sayoni S., thanks for getting back. Really appreciated. I'll get in touch "privately".

Maybe, if we can work out how to describe things clearly (and simply) we can post something back onto this thread in case it would help others. I'd be happy to put my shoulder to the wheel regarding that.

(This is the kind of thing that makes RG shine.)

But first things first I'll contact you!

Robt.

Kindly allow me to add to Daniel's query. It might be useful to provide additional information about the construct of interest. That is, eight dimensions of a single construct [such as depression], or "8 unique constructs" [such as depression, anxiety, anger...]? Does the structure of the instrument change, after removing a single item? It also helps to discuss the factor analytic method undertaken.

Robert Trevethan ...That's very generous of you to say that, Robt. So, on a quick note, AVE is an important aspect to establish the convergent validity of a scale, used to measure a construct. Say, we have a construct 'Z'. We perform EFA in SPSS to determine its factor structure. As a result, we obtain three factors, Za, Zb, Zc, each having three items. This EFA analysis provides the dimensionality of the construct 'Z'. Now, to determine the AVE, we perform CFA in SPSS AMOS. The CFA establishes the measurement model of the construct 'Z', along with providing the item loadings (plus error terms that are not obtained through EFA).

For example,

in Za, we have three items with loadings: Za1=0.80, Za2=0.75, and Za3=0.84

So, AVE score of Za= Sum of square of individual item loading/no. of items, =[(0.80)2 +(0.75)2+(0.84)2]/3=(0.64+0.56+0.71)/3=1.91/3=0.63. The score, 0.63, indicates that items Za1, Za2, and Za3 , converge together, and able to explain 63% variance of the construct. Remaining 37% is noise/error.

Similarly, AVE for other factors (Zb and Zc) can be calculated.

The thumb rule is, the AVE score needs to be greater than 0.50 to confirm the convergent validity.

The above explanation is very generic though. There are so many important aspects that go around in establishing psychometric properties of a scale. For example, the correlation between the factors (for determining discriminant validity), types of a construct (primary v/s secondary), type of tool used (eg., R gives direct output) and so on. I'll happy to elaborate more in further communications.

Hope I make sense, and this thread is helpful.

Many thanks

Sayoni

Sayoni S., thank you enormously for getting back, and for making your explanation available to more people than just me.

NOW: I do, indeed, need to confess that I didn't know what the AVE was (it wasn't something I was familiar with but that was masquerading under a different name).

I have a question, however: Are the item loadings those that are obtained through CFA or through EFA?

I'd like to try it out with some of the data I've been working on.

Thanks again. Ooooh, but near the end of your post immediately above, do you mean threat or thread? I really hope the latter. That's what I perceive from the whole tone of your post . :-) :-)