Important: A jumble of items is not a psychometric scale. A psychometric scale is a collection of items that all assess *the same construct*.

Things to watch out for:

• Multiple-construct scales (example: The Big Five Inventory, which has items assessing five separate personality traits). If you have a scale like this, you need to identify the items that measure your specific construct. Those are your scale, not the others. Never average things together that aren't the same kind of thing.
• A bunch of questions that aren't a scale at all. This happens when you make something up yourself (making scales can be *hard!*), or when you find someone's scale on their janky website tucked between animated GIFs from the 1990s. Search the scholarly literature. Get your scales from peer-reviewed publications. Even then, be a bit skeptical.
• Scales with different kinds of response options. There are scales with some items on a 1-5 “strongly disagree” to “strongly agree” scale, then a few items that are yes/no. The instructions in this video assume that all items are measured with the same response options (confusingly, these options are also called a “scale”. Yes, we have issues).

Make sure you have

• The scale items, word-for-word
• The response options (a.k.a. the “response scale”)
• The scoring instructions (if you can't find these, and if the items all have the same response options, just do the average of all items)
• A list of all reverse-coded items. Some scales don't have any (yay! It's easier to score them!), but many have a few. You need to know exactly which these are, and be able to identify their response columns in your dataset.

Here's the video. It is 24 minutes long because this is important and has some fussy details.