Determination of sample size: Brains were prepared for imaging in batches of 5–10. In initial batches, we assessed the variability of the manipulation, for example if we were trying to change Kenyon cell number, we looked at how variable the size of the Kenyon cell population was following the manipulation. We used this variability to determine how many batches to analyze so as to obtain enough informative samples. To avoid introducing statistical bias, we did not analyze the functional consequence of the manipulation until after completing all batches; for example, if the manipulation was intended to alter Kenyon cell claw numbers, we did not quantify odor responses until after completing all samples. Similarly, we did not measure the effect on other cell types (such as assessing projection neuron bouton phenotypes) until after completion of all samples. Genotypes or conditions being compared with one another were always prepared for staining together and imaged interspersed with one another to equalize batch effects except in occasional cases that have been highlighted in methods above. Since our developmental manipulations seemed to affect the mushroom body in each hemisphere independently and variably, we treated each hemisphere as an independent sample in our staining and functional imaging. In the case of hydroxyurea-treated animals, if one hemisphere was affected severely, the other one is likely affected to a similar extent but not necessarily equally, e.g. we observed KC clone counts of “[1,0]” or “[3,2]” but never “[4,0]”.
Criteria for exclusion, treatment of outliers: We excluded from analysis samples with overt physical damage to the cells or structures being measured. In figures and analyses, we treated outliers the same way as other data points. For in vivo functional imaging experiments, full criteria for inclusion and exclusion of each sample are discussed above under “Analysis of KC somatic odor responses” and “Analysis of MBON odor responses”.
Statistical tests applied are mentioned in each figure legend along with the p-value significance. To communicate our findings in the simplest and most complete way, we have displayed each data point for each sample to allow readers to assess effect size and significance directly. When sample size could be determined from the figures, we did not explicitly state it in the figure legends.
All statistical analysis were performed in GraphPad Prism, Excel, or R.