0
$\begingroup$

Just wanted to ask everyone on here if I got the right idea. I'm doing research on algal cultures and I have data from 6 treatments which have three replicates each. A friend of mine did similar research before and he did a one way anova test on his data.

Just wanted to clarify if my thought process is correct. The treatments are the independent variables, and the anova is primarily done to see whether there is a significant difference in the means between treatments?

And regarding SPSS, does the program provide the superscripts for the Tukey's HSD? Or do you have to so it yourself?

$\endgroup$
  • $\begingroup$ Questions on how to do this in a particular programming package (e.g. SPSS) is off topic here. $\endgroup$ – Michael R. Chernick Apr 13 '17 at 16:44
  • $\begingroup$ Chances are there is a single Treatment variable that has 6 levels or categories: treatment A, B, C, etc. The anova might be done to test for significant differences as you mention, and/or to specify the means and variabilities of the outcome at each treatment, and/or to see to what extent the Treatment can account for the variance in the outcome. Unclear to me what you mean by "superscripts." $\endgroup$ – rolando2 Apr 13 '17 at 16:46
0
$\begingroup$

The larger question when using ANOVA is whether or not the variances of the outcome variable for the 6 groups are equal, called "homoscedasticity". You only have 3 objects per group, which cannot provide reliable estimates of variance. If you did have larger sample sizes per group, you could employ the W-test. Thus, you might use SPSS's Kruskal-Wallis rank-based non-parametric test instead, which is not based on means and variances. Either that, or convert your observations into into ranks, and then run ANOVA on the ranks.

Tukey and Scheffe in SPSS are mostly for all possible comparisons of means.

$\endgroup$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.