ANOVA or nonparametric test

Hi. I am trying to analyse the results of my experiment for my diploma. I had 10 different plant medium, 2 different types of plant explants with 6 replications. I cultured plant explants in vitro for 6 weeks, after that time i counted new shoots, which developed from explants. The number of new shoots differentiated due to different plant medium and type of explant. In most cases there wasn't any new shoot or just 1 per 6 replications; but in one case there were 13 new shoots per 6 replications (one medium really stood out).

I wanted to carry out ANOVA test, but the Levene's test failed (the upper picture). Then i transformed the data using log10 + 1 (the bottom picture), but Levene's test failed again. Should i try any other transformation or is the only solution using a nonparametric test?

I have one more question. Can i carry out ANOVA if Levene's test go through but the diagnostic plot shows that residuals are not normally distributed?

Best regards,


Try Poisson regression instead with "plant medium" as explanatory factor. It is similar to anova but uses the Poisson distribution (for number of shoots) instead of the normal distribution. (It uses the dependent variable "shoots" given the explanatory factor, thus it corresponds to the residuals.)
Hi again. Below is the output of the Poisson regressin (i hope i did it right). But now i don't exactly know what these numbers are saying. Ok, i know that Gojišče [T.9] shows statistical significance (that's the plant medium which stood out).
How do i know if this model is ok? What does the value of AIC mean? I miss p value or at least R squared :D

glm(formula = Poganjki ~ Gojišče, family = poisson, data = advreg)
Deviance Residuals:
Min 1Q Median 3Q Max
-1.7795 -0.7071 -0.5774 0.3335 2.2344
Estimate Std. Error z value Pr(>|z|)
(Intercept) -1.792e+00 7.071e-01 -2.534 0.01128 *
Gojišče[T.2] 6.931e-01 8.660e-01 0.800 0.42349
Gojišče[T.3] 1.253e+00 8.018e-01 1.562 0.11818
Gojišče[T.4] 1.386e+00 7.906e-01 1.754 0.07951 .
Gojišče[T.5] 1.485e-15 1.000e+00 0.000 1.00000
Gojišče[T.6] -6.931e-01 1.225e+00 -0.566 0.57143
Gojišče[T.7] -6.611e-17 1.000e+00 0.000 1.00000
Gojišče[T.8] 4.055e-01 9.129e-01 0.444 0.65692
Gojišče[T.9] 2.251e+00 7.434e-01 3.028 0.00246 **
Gojišče[T.10] 4.055e-01 9.129e-01 0.444 0.65692
Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
(Dispersion parameter for poisson family taken to be 1)
Null deviance: 140.97 on 119 degrees of freedom
Residual deviance: 102.16 on 110 degrees of freedom
AIC: 200.94
Number of Fisher Scoring iterations: 6


Less is more. Stay pure. Stay poor.
AIC is a value for model fit. If you ran multiple models this value could be contrasted between them with model selection based/influenced by lower AIC values.

Of note, it has been awhile since I have ran a Poisson model, but I believe your output now is not similar to that of linear regression. It may now be the relative risk.