Methods
Growth measurements were taken from each of the 192 lodgepole pine seedlings at the end of the two month long experiment.
The seedlings were a year old when the experiment began. All lodgepole pine seedlings had set bud (finished height growth) prior to measurements being taken.
|
Picture 1. Sketch of greenhouse containing lodgepole pine seedling experiment. Shows eight blocks, and layout of room.
These measurements included height, water use efficiency and stomatal conductance. Height was measured using a regular ruler, and water use efficiency and stomatal conductance were measured using a LiCor 6400XT Portable Photosynthesis system.
|
The experiment was organized in a complete random block design with four treatment groups. 1: well-watered and not inoculated with fungus, 2: well-watered and inoculated, 3: drought and not inoculate, 4: drought and inoculated. There were eight blocks, and two replicates of each treatment of each breeding class per block for a total of 24 seedlings per block and 192 seedlings in the experiment.
|
Picture 2. Outline of experimental design. A total off eight blocks, each with 24 seedlings, comprised of elite, improved and unimproved seedlings, each with the four treatment groups, and two seedlings of each.
Data Exploration and Statistical Analysis Methods:
I looked at scatterplots, histograms and boxplots of my raw data. I felt that box plots were the most space effective way to display my data. This also allowed me to see outliers and be able to check the raw data to ensure those values were in fact correct.
I decided on a Classification and Regression Tree Analysis (CART) because this was a straightforward way to approach my research question: did breeding class still make the biggest difference in terms of seedling growth, even when the seedlings were subjected to two environmental stressors? I could look to see which factors (breeding class, treatment group or even block) most significantly impacted the differences in growth that were observed. This technique splits the data into groups, or branches of the tree, in such a way as to minimize variance in the response group. The first split of the tree is the factor that explains the most variance in the dataset.
I used three One-Way Analysis of Variance tests to look at the significance of the three growth measurements: height, water use efficiency and stomatal conductance. I tested the assumptions for normality and homogeneity of variances and concluded that my data was normal enough for an ANOVA.
I looked at scatterplots, histograms and boxplots of my raw data. I felt that box plots were the most space effective way to display my data. This also allowed me to see outliers and be able to check the raw data to ensure those values were in fact correct.
I decided on a Classification and Regression Tree Analysis (CART) because this was a straightforward way to approach my research question: did breeding class still make the biggest difference in terms of seedling growth, even when the seedlings were subjected to two environmental stressors? I could look to see which factors (breeding class, treatment group or even block) most significantly impacted the differences in growth that were observed. This technique splits the data into groups, or branches of the tree, in such a way as to minimize variance in the response group. The first split of the tree is the factor that explains the most variance in the dataset.
I used three One-Way Analysis of Variance tests to look at the significance of the three growth measurements: height, water use efficiency and stomatal conductance. I tested the assumptions for normality and homogeneity of variances and concluded that my data was normal enough for an ANOVA.