1
$\begingroup$

I have 20 individuals who memorised 5 lists of information. They immediately recalled the information, then at the end of the day either received an interventional 'cue' or nothing prior to recalling the information again. This was performed on two separate days with a different set of lists (i.e., this was a randomised crossover study).

I want to see if there is a statistically significant difference in the information recalled using the intervention or not. Would a sign ranked test be appropriate just grouping the information and testing intervention vs non intervention for changes in information recalled? Or would an ANOVA be more appropriate?

I have Stata if that helps. example of headings used for workings

Participant Number, intervention, time, case, correct, percentage correct,

example info

Number intventn time case correct  percentage correct
        1   2   1   6   21      26.92307692
        1   2   1   7   14      25.45454545
        1   2   1   8   16      21.62162162
        1   2   1   9   18      35.29411765
        1   2   1   10  15      19.23076923
        1   1   1   1   13      16.04938272
        1   1   1   2   6       10.71428571
        1   1   1   3   4       5.194805195
        1   1   1   4   7       14.28571429
        1   1   1   5   6        8
        3   2   1   6   37      47.43589744
        3   2   1   7   28      50.90909091
        3   2   1   8   31      41.89189189
        3   2   1   9   24      47.05882353
        3   2   1   10  32      41.02564103
Improve this question
$\endgroup$
8
  • $\begingroup$ What is the response exactly? Is it number of items recalled out of some total number of items? Can you clarify the nature of the lists? $\endgroup$
    – gung - Reinstate Monica
    Commented Jan 9, 2015 at 19:59
  • $\begingroup$ its the number of items recalled. the lists are items of information, for example it i describe pain, i describe its site, position severity, etc, therefore site, position and severity would each receive a mark. $\endgroup$
    – ianf
    Commented Jan 11, 2015 at 13:51
  • $\begingroup$ Num interventn time case correct percentage correct 1 2 1 6 21 26.92307692 1 2 1 7 14 25.45454545 1 2 1 8 16 21.62162162 1 2 1 9 18 35.29411765 1 2 1 10 15 19.23076923 1 1 1 1 13 16.04938272 1 1 1 2 6 10.71428571 1 1 1 3 4 5.194805195 1 1 1 4 7 14.28571429 1 1 1 5 6 8 $\endgroup$
    – ianf
    Commented Jan 11, 2015 at 15:49
  • $\begingroup$ sorry went wrong above $\endgroup$
    – ianf
    Commented Jan 11, 2015 at 16:05
  • $\begingroup$ I've posted an example of the data on next post. Its a crossover so the cases he recalled were 1-5 and 6-10 on different days intervention or no intervention. The amount of info in each case was different but total number of items the same over 5 cases. For the written version Shap wilk <0.05 for correct & percent corr for intervention group so I did Kwallis (can i do that as its grouping all the results of cases together?). my stata command was 'kwallis percentage correct, by(intervention)' $\endgroup$
    – ianf
    Commented Jan 11, 2015 at 16:05

1 Answer 1

Reset to default
0
$\begingroup$

Since you have a randomized crossover design, you should be looking at intervention, sequence, and potentially list as independent variables. My hunch is that your data will be pretty normally distributed (if your dependent variable is number of items recalled) and ANOVA will be fine, but you can check that assumption and use non-parametric approach if warranted.

Improve this answer
$\endgroup$
1
  • $\begingroup$ If the DV is number of items recalled out of a total number of items, it will be distributed as a binomial, not normally. $\endgroup$
    – gung - Reinstate Monica
    Commented Jan 9, 2015 at 20:50

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.