 |
 |
Do you have this item for some people? How many? What is the R-squared for this item when you regress it on all the other items for people who have the whole questionnaire? If it's high (say, > 0.6), I'd run a single imputation. If it's not, I'd consider running multiple imputation, but I'd probably just do mean score of items on the same subscale. The advantage you have is that if this item is highly correlated with the other items, it is easy to predict what the score would be. If the item is not highly correlated, it just adds random noise anyway, so you can use mean substitution and it won't hurt. Jeremy On 30 June 2014 08:06, Ciara gill <[log in to unmask]> wrote: > I have had the most unfortunate experience with my DClinPsy Research. I > was conducting a simple online questionnaire project with adolescents. > However due to lack of access to computers a number of schools opted to > print the questionnaires. Unfortunately something to do with the formatting > from Bristol Survey resulted in the last question Q26 of the > Self-Compassion Measure (attached) not being printed. >