Nested SPSS
To run a nested SPSS we are investigating whether a whole group of genotypes have an effect. For example, is there a general effect of controls v a set of mutations v rescues. Sadly, this can’t be done from the SPSS menus.
Lay out the data in the usual SPSS way, like this, with the data column, the genotype column, and a class column (control, mutant, rescue):
Chose File | Open | Syntax and select nested.sps If you have done something other than count boutons, edit the text to replace bouton by the name of your measurement (e.g. EJP).
This will do either a conventional ANOVA or the nested ANOVA. The difference lies in the last line of each section, with the top half running a conventional ANOVA “ /DESIGN=genotype class genotype*class.” while the second block does the nested ANOVA by virtue of the line “ /DESIGN=class class(genotype). “ Note the brackets!.
To run the nested ANOVA, click in the second block and choose Run | Selection (or press the big green arrow).
continued overleaf…
You should get output including a table like this:
Dependent Variable:bouton /
Source / Type III Sum of Squares / df / Mean Square / F / Sig. /
Corrected Model / 164574.605a / 14 / 11755.329 / 24.722 / .000 /
Intercept / 4085301.959 / 1 / 4085301.959 / 8591.655 / .000 /
class / 139796.604 / 4 / 34949.151 / 73.500 / .000 /
class(genotype) / 16223.243 / 10 / 1622.324 / 3.412 / .000 /
Error / 169752.250 / 357 / 475.496 /
Total / 4560540.249 / 372 /
Corrected Total / 334326.855 / 371 /
a. R Squared = .492 (Adjusted R Squared = .472)
The line highlighted in red shows if there is a group effect (control v mutant v rescue). The line highlighted in blue shows if, within the groups, there is significant variation, ie if the P value > 0.05, the classes are homogenous (They aren’t in this data set – the controls vary quite a lot!).
And no, you can’t do a post-hoc test for nested ANOVA.
Ask me about the output! and other possible tests!
C. J. H. Elliott,
January 21, 2011