Can I perform (run) EFA four times? one by one after deleting the variables with low communalities?

thank you for your asking. I will try to explain :)

Measures and Materials

I have two questionnaires (first QUSE and second QCOBESE (that I made and checked during the preliminary study for my project) and I would like to analyse using factor analysis (independent variables).

I have questions about personal integration in the host country as well (dependent variables).

Responses to The 25-item QUSE and The 40-item QCOBESE were analysed by factor analysis. To identify the factor structure of the scale, principal axis factoring (PAF) and oblimin rotation (OR) in SPSS (package V29 (IBM SPSS Statictics).

First one.

EFA (1-3) were performed on all 25 items extracted from relative scales using the sample. Of the 283 respondents, 10 were missing 1 or 2 item responses for a total of 25 missing data points.

We examine EFA1 QUSE with 25 items, which combined into 2 factors that explain 35.3% of total variance (The Kaiser-Meyer-Olkin (KMO)=0.86 𝜒²(300) = 2266.79, p < .001, F1= 27.1%, F2=8.2%). The two-factor solution of the structure matrix was next examined for adequacy. We found 8 items that had less than 0.25 commonalities (h2 from 0.14 to 0.25). EFA2 QUSE had 17 remaining items, which were grouped into 2 factors, 44.8% of total variance (The Kaiser-Meyer-Olkin (KMO)=0.86 𝜒²(136) = 1579.81, p < .001, F1= 33.3%, F2=11.5%. When we analysed the structure matrix we found 1 item which had low factor loading. Afterwards, 16 items remained.

Final EFA3 QUSE had 16 remaining items, which were grouped into 2 factors, 46% of total variance (The Kaiser-Meyer-Olkin (KMO)=0.85 𝜒²(120) = 1481.52, p < .001, F1= 33.8%, F2=12.2%. Given the number of participants in this study, structure coefficients ≥.37 were considered salient (i.e., both practically and statistically significant, as per Norman & Streiner, 2014). Complex loadings that were salient on more than one factor were rejected to honour simple structures (Thurstone, 1947). Factors with a minimum of three salient structure coefficients, internal consistency reliability ≥.70, and that were theoretically meaningful were considered adequate. The two factors suggested by EFA3 QUSE were named ”.....”.

Second one

Four sets (EFA 1- 4) QCOBESE using the sample 234 (from 283) participants of rotated factors were compared. The EFA1 had 40 items, which combined into three factors, all items adapted from preliminary studies C and D. After the Empirical Kaiser criterion and screen plot, a three-factor structure would be an appropriate factor solution. Three factors that explain 35.8% of total variance (F1 = 17.4%, F2 = 12.4%; F3 = 6%,The Kaiser-Meyer-Olkin (KMO)=0.76 𝜒²(780) = 2304.58, p < .001) The factor loadings after oblimin rotation showed that 16 items had low factor loadings and low communalities were not included in any factor and we had removed them (see in Table 8). A factor analysis EFA2 yielded the expected 3 factors, which together accounted for 47.1 per cent of the variance in the 24-item set F1=24.5%, F2 =15.3%; F3= 7.3%,KMO was 0,82. and Barlett’s test of sphericity was significant (χ2 = 1573.42, df = 276, p < .001). After analysing the commonalities, the 24th items of EFA2 and the factor correlation matrix, we found out that 1 and 3 factors had strong significant negative correlation (-0.344 p≤0.01). We removed the next 4 statements with low communalities (less than 0.25). The EFA3 had 20 items, which combined into two factors. Two factors that explain 44.9% of the total variance (F1 = 28.9%, F2 = 16%). The Kaiser-Meyer-Olkin (KMO)=0.85 𝜒²(190) = 1419.18, p < .001). We checked the value of the commonality indicator for it and removed 2 items again. The final fourth EFA had 18 remaining items.

Regarding PAF, a final solution EFA4 with two components (F1 and F2) was obtained in oblimin rotation method. Two factors accounted for 47.1 per cent of the variance in the 18-item set F1=30%, F2 =17.1%. KMO was 0,86. and Barlett’s test of sphericity was significant (χ2 = 1269.17, df = 153, p < .001). These two components whose eigenvalues were greater than one emerged. are shown in Table 7. All the factor loadings were above or close to 0.45. The structure factor matrix in component 1 comprised 11 items, the second component comprised 7 items. The two factors suggested by EFA4 were named ..... respectively.

Hypothesis 1: Participants with a high level of QCOBESE feel less QUSE

Hypothesis 2: QUSE of the participants adversely affects their exercise of personal integration