Which statistical test can I use?
Hello first of all,
I'm currently analyzing a survey for my bachelor's thesis. I have the following hypothesis: "It's more likely that product A will be offered than product B." To test this hypothesis, I would like to use two questions and their results from my survey. I wanted to use the Wilcoxon signed-rank test, since the same participants answered the questions and I used a 5-point Likert scale (ordinal scale). However, I noticed that a Wilcoxon signed-rank test requires two groups.
Now to the real question: Can I still use this test, or do I need to take a different one? And if I do need to take a different one, which one should I use?
Thank you very much in advance and kind regards.
(2 votes)
Loading...
There is the Wilcoxon ranking test (2 groups) and the Wilcoxon sign ranking test (dependent measurements). The latter requires interval-scale data. Your rating list (one item is NB No Likert scales consisting of several items) is strictly ordinal as you write yourself. The sign test is suitable for this.
See similar problems by search engine Statistics of decision tree, that is sometimes helpful.
Can I also use the sign-travel test if I only asked a group of people? I don’t know what the groups are. I just want to know what the test persons have rated better.
Thank you for your answer
The Rank sum test is for comparing two groups, but this is not about. Conversely, Indication number is for comparing dependent measurements (2 measurements on the same group of persons). However, it requires interval scales. The sign test is suitable for dependent measurements at ordinate scale level.
Hello @Machma200 can I define one question and the other question as a group? Or it is imperative that I make two surveys with the same people or a survey with two groups. I would prefer to test the two questions from one survey using the Wilcoxon sign-Rang test.
2 questions are not 2 groups. The sign-Rang test is for 1 group that has been interrogated twice, but requires interval scale. Use ordinate scale, then sign test.
That’s the right test. 2 groups are also called paired groups here, they are not independent, and precisely these pairwise correspondence is one of the prerequisites for the Wilcoxon sign-Rang test. For more details see my comment on LeBonyt. However, you do not meet all the requirements for the Wilcoxon sign-Rang test (stability), and the simple sign test, as Machma2000 writes, is the alternative here. However, this takes into account only the direction of the difference, but not when product A and product B are rated better at the same time, but at A the difference to B is usually higher than vice versa (approximately A vs. B 1:5, B vs. A 2:3).
If you can confirm your hypothesis with both tests, the better
Are 2 groups also meant groups of people? I have only made a survey or asked only one group of people and would like to test which product was most rated better. Can I still use the sign-Rang test?
The Wilcoxon Sign Rang Test is a non-parametric statistical test that can be used to test whether two associated samples have significant differences in their averages. In this case, the test would be used to test whether the mean values of variable A and variable B differ significantly.
In order to carry out the test, the answers of the subjects to the two questions must first be converted into a range system. This can be done by assigning the highest ranking (1) for the highest response and the lowest ranking (5) for the lowest response. Subsequently, the gears for each variable are added in order to obtain the sum number numbers for variable A and variable B.
The Wilcoxon sign-Rang test then calculates the p value which indicates the probability that the observed differences between the mean values of the two variables occur randomly. If the p value is smaller than the significance value, then the zero hypothesis is discarded and assumed that there is a significant difference between the mean values of the two variables.
In this case, the test would be used to test whether the mean values of variable A and variable B differ significantly. If the p value is smaller than the significance value, then the zero hypothesis would be discarded and assumed that the variable A is offered more frequently than the variable B. However, if the p value is greater than the significance value, then the zero hypothesis would not be discarded and no significant difference between the mean values of the two variables would be determined.
It is important to note that the Wilcoxon sign-Rang test is a non-parametric statistical test, which means that no assumptions have to be made about the distribution of the data. This makes the test a good choice for use with data that are not distributed normally.
Can you please give feedback on whether Bard’s written crap or something.
There are a lot of detail errors in it, but the thick knocker is the Wilcoxon sign ranking test and Wilcoxon ranking test.
Thanks for your feedback.
You don’t know Bard? That may not be so bad that you don’t know him. The whole article was written by him. Conclusion : Bard is a salmon number and therefore does not dedicate anything to complex questions. Thanks for the trouble.