I would like to ask a question about two way anova using SPSS. I want to compare 13 grape varietes (Independent Variable) for some physicochemical characteristics (e.g. reducing sugars which is the dependent variable) and simoultaneously according to their type (Independent Variable). Type means the colour. I used two - way anova and here is the problem. My first problem is that at the table of tests between subject effects the raw type has a (-) and the second problem is Levene's test which is significant. As for the first problem I don't know what to do to solve it. Without this Can i move on to the two - way anova procedure or it is necessary?? As for the Levene's test after hours of reading I figured out that the only thing I can do for the post hoc comparison is to use a common test such as Tukey's test and compare them using p value
Hello Christos,
Based on the anova table you furnished, your data set, as coded, has some issues. The variable "type" had only 1 level (e.g., was a constant in your data set), or had no levels. The df for varieties should have been 12, not 10.
I think the first order of business should be to check your data set, and ensure that the values for the respective variables are appropriately named and coded for the cases. Then, you can evaluate any and all assumption conditions usually made for anova and go from there. SPSS does offer bootstrap estimates of effects and error terms, which are more robust against violations of the normality/homogeneity conditions. As well, SPSS offers post hoc test that are robust against homogeneity concerns (e.g., Games-Howell; Tamhane).
Good luck with your work.
Dear David Morse thank you very much for your instant response.
I would like to ask you something more if you could help me. The variable type has 3 levels (3 different colours). I will upload a pic of my data in order to tell me where I am wrong so I can fix it if that is possible. Also, you believe tha I should perform two-way anova or one for each variable seperately? I'm asking because i tried that on my own but the results are different. Finally to one-way anova the post hoc tests you mentioned before are correct. In two way anova I can't use them. So in case that Levene's test is significant what can I do? I'm attaching a photo to show briefly how i have entered my data.
Kindly check https://statistics.laerd.com/spss-tutorials/two-way-anova-using-spss-statistics.php
Also check https://www.spss-tutorials.com/spss-two-way-anova-basics-tutorial/
Before using ANOVA for data you must test data distribution normality, additive effects, homogeneity of variance, these assumptions must be available in your data first before using anova, if there are problem in one or more assumptions, then you can use transformations (such as log, sqrt, ...) to change distribution to normal distribution. If all attempts failed then you can use non-parametric tests (such as Kruskal-Wallis test) rather than anova
Hello again, Christos,
Your data are entered appropriately. What I'd suggest is that you run a crosstabulation table on the two factors, and note: (a) the total number of cells with any cases; and (b) the number of cells with more than 1 case. This should help to explain the apparent problem when you tried the two-way analysis (which is, by the way, the appropriate design given what you've said about your data set).
Good luck with your work.
- Badges
- Science topic
- Similar topics
- Mathematical Sciences
- Statistics
- Statistical Software
- SPSS