Flashcards in BIO 330 Deck (379)

Loading flashcards...

151

## Hypothesis test steps

###
State Ho, Ha

calculate test statistic

determine critical value of null distribution (or P-value)

compare tests statistic to critical value (or P-value to sig. level)

evaluate Ho using alpha

152

## why use alpha = 0.05

### balances Type I error and Type II error

153

## why are Type I and II errors conceptual

### we don't know whether or not Ho is actually true

154

## paired t-test is a type of

### blocking

155

## where does pseudoreplication happen/become a problem

### data analysis stage, doesn't happen at data collection stage (subsamples)

156

## ANOVA maintains

### P[Type I Error] = alpha

157

## ANOVA, Y bar

### grand mean, main horizontal line, test for differences between grand mean and group means

158

## ANOVA, Ho: F-ratio =

### ~1

159

## ANOVA, if Ho is true, MSerror

### = MS groups; same variation within and btw goups

160

## ANOVA, MSgroup > MSerror

### more variation between groups than within

161

## ANOVA, test statistic

###
F-distribution, F_0.05,(1),MSgroup DF, MSerror DF = critical value

compare critical value to F-ratio

this is a one sided distribution we are looking for whether F-ratio is bigger than critical value (strictly)

162

## ANOVA, F-ratio > F-critical

### Reject Ho.. at least one group mean is different than the others

163

## ANOVA, quantifying variation resulting from "treatment effect"

###
R^2 = SSgroups/SStotal

R^2 [0,1]

164

## ANOVA, high R^2

### more of the variation can be explained by the treatment, usually want at least 0.5

165

## ANOVA, R^2 = 0.43

### 43% of total variation is explained by differences in treatment

166

## ANOVA, R^2 = low values

### noisy data

167

## ANOVA assumptions

###
Random samples from populations

Variable is normally distributed in each k population

Equal variance in all k populations

168

## ANOVA unmet assumptions

###
large n, similar variances-- ignore

variances very different-- transform

non-parametric-- Kruskal-Wallis

169

## ANOVA, which group(s) were different

### Planned or Unplanned comparison of means

170

## Planned comparisons of means (ANOVA)

### comparison between means planned during study design, before data is obtained; for comparing ONE group w/ control (only 2 means); not common

171

## Unplanned comparisons of means (ANOVA)

### comparisons to determine differences between all pairs of mean; more common; controls Type I error

172

## Planned comparison calculations (ANOVA)

###
like a 2-sample t-test

test statistic: t =(Ybar1 - Ybar2)/SE

SE= √ MSerror (1/n1 + 1/n2)

note that we use error mean square instead of pooled variance (as in a normal t-test)

df = N-k

t critical= t0.05(2), df

173

## Unplanned comparison of means (ANOVA)

### Tukey-Kramer

174

## why do you need to know what kind of data you have

### determines what kind of statistical test you an do

175

## left skew

###
mean < median

skew 'pulls' mean in direction of skew

176

## C.I. notation

### 95% CI: a < µ < b (units)

177

## accept null hypothesis

###
NEVER!!!

only REJECT or FAIL TO REJECT

178

## why do we choose alpha = 0.05

### it balances TIE and TIIE which are actually conceptual, since we don't know if Ho is actually true or not

179

## standard error or estimate

### standard deviation of its sampling distribution; measures precision of the estimate

180