Survey results statistically significant or not
- Thread starter amr344
- Start date
Did you have in mind beforehand to compare that one particulkar response with the other responses? Or was this particular response chosen after inspection of the data?
Apart from that: the data are repeated measures, i.e. any of the 54 participants chose yes or no for each of the 14 stimuli (plus "other).
To analyse such data appropriately you'll need the individual behaviour of each participant. In that case, you can compare the responses to arbeitsrechtshelden.de with the responses to e.g. anwaltheute.de (the next frequent choice) using McNemar's test.
With kind regards
Karabiner
Apart from that: the data are repeated measures, i.e. any of the 54 participants chose yes or no for each of the 14 stimuli (plus "other).
To analyse such data appropriately you'll need the individual behaviour of each participant. In that case, you can compare the responses to arbeitsrechtshelden.de with the responses to e.g. anwaltheute.de (the next frequent choice) using McNemar's test.
With kind regards
Karabiner
Did you have in mind beforehand to compare that one particulkar response with the other responses? Or was this particular response chosen after inspection of the data?
Apart from that: the data are repeated measures, i.e. any of the 54 participants chose yes or no for each of the 14 stimuli (plus "other).
To analyse such data appropriately you'll need the individual behaviour of each participant. In that case, you can compare the responses to arbeitsrechtshelden.de with the responses to anwaltheute.de (the next frequent choice) using McNemar's test.
If you do not have the individual respones, then you can test on basis of an artificial scenario. That is, you assume that all of those who said "yes" to anwaltheute.de also said "yes" to arbeitsrechtshelden.de.
With kind regards
Karabiner
Apart from that: the data are repeated measures, i.e. any of the 54 participants chose yes or no for each of the 14 stimuli (plus "other).
To analyse such data appropriately you'll need the individual behaviour of each participant. In that case, you can compare the responses to arbeitsrechtshelden.de with the responses to anwaltheute.de (the next frequent choice) using McNemar's test.
If you do not have the individual respones, then you can test on basis of an artificial scenario. That is, you assume that all of those who said "yes" to anwaltheute.de also said "yes" to arbeitsrechtshelden.de.
With kind regards
Karabiner
Karabiner has some good advice, as usual.
I am assuming that you looked at the data after it was in and chose the largest, then asked is it significant?
A less conventional possibility could be a randomization test. You assume that responses are at random. Allocate the total number of positive responses to the subjects/options at random and find the total for each option. Find the maximum total. Write it down. Repeat many times to give you a sampling distribution of the maximum assuming no pattern to the responses. See where 34 fits in this distribution. If less than 5% of your sampling distribution is more than 34 then 34 is significant. kat
I am assuming that you looked at the data after it was in and chose the largest, then asked is it significant?
A less conventional possibility could be a randomization test. You assume that responses are at random. Allocate the total number of positive responses to the subjects/options at random and find the total for each option. Find the maximum total. Write it down. Repeat many times to give you a sampling distribution of the maximum assuming no pattern to the responses. See where 34 fits in this distribution. If less than 5% of your sampling distribution is more than 34 then 34 is significant. kat