• Exam 2 Review• Institutional Review Board⁃ Reviews studies to protect subjects from possible harm⁃ Does NOT decide whether study produces valuable information⁃ At least five members⁃ Scientist, non-scientist, one unaffiliated• Informed consent before data collected⁃ Informed in advance of any risk of harm⁃ usually in writing• Confidentiality (no individual data, only summaries)• Anonymity (subjects' names not known - makes follow-up difficult)• Clinical Trials⁃ Phase 1 - small group (20-80) evaluate safety, determine safe dosage⁃ Phase 2- larger group (100-300), evaluate effectiveness, further evaluate safety⁃ Phase 3, large groups (1,000-3,000) confirm effectiveness⁃ Phase 4 - post marketing status• Valid measure of a property - how a rate can be more valid measure than a count• Predictive validity - measure can be used to predict success on tasks related to property measured• Bias - systematically overstates or understates true value⁃ Reduce bias by getting a better instrument• Reliability - result is repeatable (random error is small)⁃ Increase reliability by averaging several measurements• Percent Change = (current - previous) / previous x 100⁃ positive result -> percent increase⁃ negative result -> percent decrease (can't have a decrease of >100%)• Types of variables:⁃ Categorical -⁃ Nominal - categories, not ordered⁃ Ordinal - categories, ordered⁃ Ex: months⁃ Quantitative⁃ Discrete - countable⁃ Ex: number of friends, number of classes⁃ Continuous - can be measured⁃ Ex: time, weight, height• Interquartile range = Q3 - Q1• Outliers: <Q1-1.5(IQR) or >Q3+1.5(IQR)• 5 number summary (box plots)⁃ Minimum⁃ Q1 (median of data below the distribution median)⁃ Median (Q2)⁃ Q3 (median of data above the distribution median)⁃ Maximum• Mean is strongly affected by extreme values• Standard deviation - average deviation of observations from the mean⁃ if the data is more spread out, there will be a larger standard deviation⁃ s = 0 only if there is no spread or variability in the data (all of the values in a set of data are exactly the same)• Median is NOT strongly affected by extreme
View Full Document
Unlocking...