For my master's thesis, I conducted an infection assay experiment on wheat plants in pots to test several treatments for their effectiveness to control the pathogen. I had 10 variants in total. Each variant consisted of 5 pots, where 2 of them were put in a frame and measured each day for another, rather unrelated experiment (Hyperspectral Imaging). As the pots in the frames were moved daily to a measuring chamber, lied under lights for several minutes and were fixed in the frame, we initially decided to score those pots seperately and only use the 3 other pots per variant to test the effectiveness of the treatment.
As the data contained many zeros and didn't follow a Gaussian distribution, I conducted a Kruskal-Wallis test with Wilcoxon as posthoc test. Due to only having 3 repetitions per variant now, I get very high p-values. Now I wanted to test, whether being put in the frame made a significant effect on the plants/pathogens, because if not, I can combine the scoring values of the 2 pots in the frame with the other 3 to have a total of 5 repetitions per variant. For this, I plan to implement a factor called 'frame' with values 1/2 (yes/no). However, I don't know, which test to conduct here to evaluate the effect.
Do I have to conduct a confirmative factor analysis?
Thanks in advance!
As far as I can tell this makes no sense. What is the research question here?
David Booth
Hello, what's the nature of the score, is it a boolean 1/0 or a réel ?
https://www.sciencebuddies.org/science-fair-projects/competitions/experimental-design-increasing-signal-to-noise
David Eugene Booth
On the one hand, we wanted to determine the effect of different biological control agents on the pathogen by inoculating the plant with both. Then, we determined the incidence of infection for each variant by visual scoring.
On the other hand, we also wanted to test Hyperspectral Imaging for detecting the pathogen on the plants. For this, we measured 2 pots of each variant daily. That's why the conditions of these 2 pots differed with the conditions of the other 3 pots per variant.
Ayoub El Bendali
The results are the incidence and severity of the pathogen's symptoms on the plants. Thus, the values are percentages.
Hello again, I think you don't have enough data to perform statistical testing on single variants to compare them.
What i guess you must do for comparing the pots on the frame and the others, is to calculate first the mean value of your variable in each variant, in each pot type, and make a scatter plot of the results, each point represents a variant with in x the mean in the pots on the frame, and in y the mean of the variable for the three others. This way you see a certain correlation beween the mean in the pots on the frame and others. If you add a regression line, the more this line is near to y=x the more you are safe to use the 5 results as an identifier of the variant independently of whether the experiment is done on the pot or not. Also, you will have 10 points. use the R squared or the mean square error as an evaluator of your model.
Overall, since you don't have enough data, and statistical tests do not give you good results, rely mainly on plots and simpler EDA.
- Badges
- Science method