The two insomnia items (ansev1 and ansev2) do not correlate very well with the rest of the scale, and we already have another scale to measure insomnia in general. Therefore, I think it is best to occupy only the first factor without including the insomnia items.
It is common for the scales to include inverted items, where a higher score on the item reflects a lower level in the construct to be measured. For example, the scale of “tolerance towards sexual diversity” has the following items:
(tolsex1) I feel uncomfortable sharing socially with people who are of different sexual orientation from mine.
(tolsex3) Same-sex couples should not hold hands or kiss in public.
(tolsex5) I would like to see the laws changed to promote greater respect for sexual minorities.
Actually, this scale should be called intolerance of sexual diversity. A higher score on items 1, 2, 3 or 4 is associated with greater intolerance. But the fifth item is reversed: a higher score is associated with less intolerance.
Unlike the scales we reviewed earlier, these items are clearly grouped into a factor.
As it should be, item 5 has a negative relationship with the factor. This is indicated at the beginning of the factorial analysis results, in column MR1 of the standardized loadings. Furthermore, item 1 seems not to be working very well with the rest of the scale, its relationship with the general factor is clearly lower than the rest (0.25).
To calculate the alpha it is essential to add the option check.keys = TRUE when there are inverted items (otherwise the output includes a warning anyway). Adding this parameter R automatically reverses items that have a negative correlation with the rest, and adds a warning message. The inverted item has a minus sign next to the name (tolsex5-).
The alpha of the scale is good (.79), but it would be better if item 1 (.84) is eliminated.
Again, this decision has to be conceptual as well. It seems appropriate to me to eliminate it, not because it is a bad item, but because it is measuring something slightly different – an emotional reaction vs. more discursive conservative attitudes.
Before averaging the items, in any case, we have to invert the tolsex5 values! The easiest thing is to do it directly with the formula:
Now we can calculate the scale, where a higher score will reflect a lower level of tolerance towards sexual diversity. Therefore, we are going to take advantage of changing the name to “intolsex” (higher score, higher intolerance)
So far, we have seen a series of analyzes, graphs and tables to visualize the information obtained in the sample.