Dear fellow researchers,
I'm a newbie when it comes to SPSS but have manged to analyse data gathered through Likert scales reliably enough so far when comparing different groups of respondents (e.g. male/female). However, it seems I can't find the right test to check for differences within subgroups: E.g. I would like to see if there is a statistically significant difference between e.g. native/non-native speaking respondents in the group that strongly disagreed with a certain statement (i.e. gave it a 1 on the scale). I'm attaching a screenshot that may help gage the situation better.
Has anyone got experience in that area? Any tips highly appreciated. Thanks in advance,
Karin
Dear Karin,
Given that you do not have variation in your dependent variable (i.e., agreement equals 1) I suppose you may want to test whether the proportion of strongly disagreeing natives (relative to non-natives) significantly differs from a hypothesized value. You can test that using a binomial test. For application in SPSS see this link:
https://www.spss-tutorials.com/spss-binomial-test/
What you need, however, is a reasonable assumption about the distribution of rates of strong disagreement between the groups in the population. For example, you may wish to test whether the proportion of natives vs. non-natives is equal to a presumed population value of 50%. In that case, you would assume an equal distribution (or better, no systematic differences between natives and non-natives regarding the rates of strong disagreement). If you suspect an uneven distribution of the rates in the population in favor of one group (say you claim that 75% of the population of “strong disagreers” consist of natives) you can test that as well. Of course, the result is always depending on your overall sample size, i.e., small deviations of the observed value from the expected value have a greater influence in small samples.
Regards
Non-parametric test such as Sign test or
Wilcoxon matched pairs signed-ranks test. I see they are independent (gender) then use Mann-Whitney U-test.
Dear Karin,
Given that you do not have variation in your dependent variable (i.e., agreement equals 1) I suppose you may want to test whether the proportion of strongly disagreeing natives (relative to non-natives) significantly differs from a hypothesized value. You can test that using a binomial test. For application in SPSS see this link:
https://www.spss-tutorials.com/spss-binomial-test/
What you need, however, is a reasonable assumption about the distribution of rates of strong disagreement between the groups in the population. For example, you may wish to test whether the proportion of natives vs. non-natives is equal to a presumed population value of 50%. In that case, you would assume an equal distribution (or better, no systematic differences between natives and non-natives regarding the rates of strong disagreement). If you suspect an uneven distribution of the rates in the population in favor of one group (say you claim that 75% of the population of “strong disagreers” consist of natives) you can test that as well. Of course, the result is always depending on your overall sample size, i.e., small deviations of the observed value from the expected value have a greater influence in small samples.
Regards
Alexander Pabst, thank you so much for taking the time to reply, this is exactly what I've been looking for. You are officially my Held des Tages! Thank you again.
Alexander Pabst , I'm not sure a binomial test works in this case, since the objective is to compare two groups. It seems to me that a more appropriate approach is to recode the data to strong-disagree vs. not-strong-disagree. This yields a two-by-two table of counts, and a chi-square test of association, or other test can be applied. This tests specifically for a difference between two groups for strong-disagree responses.
Karin Krummenacher , I'm doubting that you really want to know if the proportion of responses for strong-disagree differ between groups. Don't you really want to know if the responses for one group are higher than the other, taking into account all possible responses? In this case you could use a Wilcoxon-Mann-Whitney test for two groups.
Béatrice Marianne Ewalds-Kvist , why do you think the values are paired between the groups? Am I missing something?
- Badges
- Science topic
- Similar topics
- Mathematical Sciences
- Statistics
- Statistical Software
- SPSS