Experimental Design

The statistical difference in data is directly correlated to its biological importance.

Statistical difference is relative to [blank_start]variation[blank_end] in the dataset, and tells you [blank_start]nothing[blank_end] about the actual biology. The biological importance of the dataset is the effect of the [blank_start]statistical difference[blank_end], and is the primary subject of the "[blank_start]discussion[blank_end]" portion of a paper.

If you perform a statistical test, and the p-test is less than 0.05, then you know that something is going on biologically.

Which of these would be examples of observational work?

Which of these were discovered/developed by Laplace and Gauss? And when?

Which of these were discovered/developed by Galton & Pearson? And when?

Which of these were discovered/developed by Gosset and Fisher? And when?

Who developed the t-test? And what pseudonym did he go under?

Number of people is an example of [blank_start]discrete[blank_end] data. Length of an elephant's trunk is an example of [blank_start]continuous[blank_end] data.

[blank_start]Accuracy[blank_end] is how close the data is to "the truth" (specifically the [blank_start]mean[blank_end]). [blank_start]Precision[blank_end] is how close multiple measurements are to each other.

Measurement Error: variation due to [blank_start]people, instruments, conditions[blank_end] Natural Inherent Variation: variation due to [blank_start]genetics, environmental history[blank_end] Treatment effect: variation due to [blank_start]treatment in experiment[blank_end]

How do you limit Measurement error?

How do you limit Natural Inherent Variation?

How do you deal with ME?

How do you deal with NIV?

Which types of tests require a normally distributed data set?

Coding a dataset changes the relationship between data poinnts.

One can often fix assumptions & issues in normality by coding and/or transforming data.

Which are examples of coding?

Which of these are examples of transforming?

Which are point estimates of the central tendency?

Which is the point estimate that best represents the central tendency in which statistical situation? skewed data: [blank_start]median[blank_end] somewhat symmetric data: [blank_start]mean[blank_end]

[blank_start]Standard Deviation[blank_end]: deviation in the sample that was measured [blank_start]Standard Error of the Mean[blank_end]: estimate in variation of mean values if you take multiple sets f samples

A random selection of individuals gives the best estimate of a population

Which of these is a balanced design?

TYPES OF RANDOMIZATION [blank_start]Random Selection with Replacement[blank_end] 1. Tag entire population 2. Random Draw 3. Replace individual in population 4. [blank_start]Probability of selection is constant[blank_end] [blank_start]Theoretically the best way.[blank_end] [blank_start]Random Selection without Replacement[blank_end] 1. Tag entire population 2. Random Draw 3. Measure Inividual 4. Do not place back into population 5. [blank_start]Probability of selection changes[blank_end] Can get around by using same measurement if drawn again [blank_start]Haphazard Selection[blank_end]: 1. Don't tag anything 2. No random draw 3. Measure Individual 4. Replacement is irrelevant 5. [blank_start]No probability of selection[blank_end] [blank_start]The worst kind of selection[blank_end], though it is still often used

As sample size (n) increases, the t-value [blank_start]decreases[blank_end], the standard error (SE) [blank_start]decreases[blank_end], and the confidence interval (CI) [blank_start]narrows[blank_end].

[blank_start]Measurement Unit[blank_end]: the object that is measured [blank_start]Experimental Unit[blank_end]: the object that is manipulated