Hey Nilanthi, a single item measure of a latent variable is very rarely a good idea. It's very difficult to assess the reliability and validity of single items, and they tend to measure only a narrow aspect of a phenomenon. There will also be a very restricted range of scores. So there is little to commend the use of a single indicator - but this does mean they have not been used. Ratings of pain are usually based on single items and some personality measures have ultra short versions that use only single items.

Maybe use your results to guide further research - whatever the single item is supposed to measure, maybe this could be better measured if you created multiple new items to tap various aspects of it.

Mark

Did you base the number of factors on eigenvalues? You might want to try other methods such as parallel analysis or setting the number of factors. You may use JAMOVI since this software gives the three methods in obtaining the number of factors in EFA.

You did not say how you conducted the EFA. Given your area of interest it sounds like you should be using something like Principal Factors Analysis rather than Principal Components, and you should be using oblique rotation to allow for correlated factors.

Nilanthi Suraweera In addition to David L Morgan suggestions, I think you should try other methods, PFA with oblique rotations or PCA with orthogonal or oblique rotation. I come across Hair et al. book "Multivariate Data Analysis," if you have 30 or more items, both PFA and PCA produce same results. My suggestion is you should try various methods of extractions and rotations however, the result should be meaningful to interpret. If not, may be you might want to consider again on your sample size and data cleaning before EFA (I think both of which you can find in the book I mentioned).

As Mark Shevlin has also said, I think 1 item in a factor may not be practical for reliability test and also if you proceed further to CFA, at least there should be 2 items in a factor (minimum 3 items seems to be more advisable). I am adding a little more again to my previous response, you must have checked things such as communalities, factor loading with respect to sample size, etc. however, just in case, you might also want to consider those things as well. Good luck....!!

Regards,

Moroo

You should consider at least 3 items. In some sources, they have also confirmed 2 items.

Nahid Darooghe Arefi The reason for not recommending the use of a two-item factor is that it is based on a single correlation, so the loadings for both items are forced to be the square root of that correlation.

In psychometrics, a measure that includes a single item can constitute a weakness relating to the conceptualization process. Starting from the idea that the item represents an indicator of the construct evaluated, we cannot be sure that with a single item we can capture the construct or the latent factor. Therefore, in the process of analyzing psychometric data, it would be impossible to estimate the metrological parameters of a measurement instrument using a single item. The same is true when the EFA indicates several latent factors of an instrument among which one factor has a single item. Generally, this factor must be deleted and the latent factors which contain at least two items must be kept. As it is recommended to estimate the reliability of the factors of a scale, it would be impossible to calculate the reliability index of a factor consisting of one item. The proof is that McDonald's Gamma method only accepts at least 3 items and Cronbach's alpha method only works when the elements are at least 2.

In your case, if you wish to keep a three-factor structure of the scale, this measure will be limited. You can remove the third 1-item factor and it all depends on the rotation methods you used during the exploratory factor analyses. Try other rotation methods in JASP, JAMOVI, SPSS, etc.