0

I would like to find out if there is a significant difference between the pre-oral reading test and post oral reading test of 18 pupils. Both the pre oral reading test composed of 5 items each on reading a word, reading a phrase and reading a sentence. What statistical analysis tool should i use? And what if the pretest scores are non-normal? Thanks!

kjetil b halvorsen
  • 77,844
Belle Perez
  • 1
  • Wilcoxon rank sum is significantly better at detecting a difference between the two samples. Pre- and Post- scores don't have to follow a normal distribution for a paired t-test, just the differences between the two scores (i.e. diff = post - pre). – Jon Dec 07 '16 at 23:58
  • Please feel free to refer to previous questions on the community as your question is quite common. Here is a good reference on the efficiency of Wilcoxon. http://stats.stackexchange.com/questions/71953/relative-efficiency-of-wilcoxon-signed-rank-in-small-samples – Jon Dec 08 '16 at 00:00
  • Check this resource. – Antoni Parellada Dec 08 '16 at 00:00
  • What do you mean by "composed of 5 items each on reading a word"? What are the data like, exactly? – gung - Reinstate Monica Dec 08 '16 at 02:01

1 Answers1

0

Probably you should add more details. How were these items coded? Are they binary items? I think you need to carry out a psychometric analysis, with the purpose of check if your five items are good in building a scale. In such case you should measure Cronbach's alpha coefficient and the correlation of each item with the five items sum. Then if you have good correlation and Cronbach alpha coefficients you can move forward comparing pre- and post-test score with paired T test or Wilcoxon depending on the normality of the score resulting from the five items. If you use R I strongly suggest the psych package. Your sample size is small, but you have few items, so you have some chance to succeed.

paoloeusebi
  • 212
  • 1
  • 7