University:
The University Of Western Ontario
Course:
Biology 2290F | Scientific Method in Biology
Academic year:

2022
Views:

375

Pages:

2
Author:

Kyuwon J.

● Population and sample ○ Population (N) ■ Set of units (individuals, objects, events) within a study ■ ex) All University staffs in Canada ○ Sample (n) ■ Subset of population in analysis ■ ex) group of 200 staff selected randomly ● Sampling frame ○ Comprehensive list + database of the sample ○ ex) list of staff for selection in university ● Parameter vs Statistics ○ Parameter ■ Fixed value on population (population mean proportions) ○ Statistics ■ Calculated values from the sample (sample mean) ● Sampling distribution ○ Distribution of given sample statistic for repeats ○ CLT as a result ■ Approaches normality as sample size increases ■ This will happen even if the population is abnormal ● Bias and Efficiency ○ Bias = difference in expected values vs true population ○ Efficiency =variance of estimator ■ Lower variance = higher efficacy ● Point estimation ○ Single best estimate of the population parameter ○ p̂ = x / n ■ Estimated population in proportion ● Internal estimation ○ Range of parameters ○ x̄ ± z* · (σ / √n) = estimation of population ■ X = sample mean ■ z* = critical z value ■ σ= population standard deviation ■ N = sample size ○ x̄ ± t* · (s / √n) = for unknown ■ T = critical t value from t-distribution (depending on sample size) ■ s= standard deviation ■ n= sample size ● ex) 95% confidence interval shows that under second trail, 95% of population will have true population mean ● Adjustment ○ Finite population correction ■ Used for sampling without replacement in finite population ● Reduces variability in estimates ● Unit cannot appear more than once = values less variable than sample with replacement ■ Sample size is non-negligible fraction of N ■ f = N / n ● Greater than 5% ● Below this, often ignored since difference is small or unrecognized ■ As n => N and FPC = 0… ● There is no uncertainty in the entire sample population ■ As n <1 ■ Design has more variance than SRS (usually due to clustering) ○ Def =1 ■ No other variance (SRS highly efficient) ○ Deff<1 ■ Improves precision (well stratified sampling) ● Cluster sampling will increase variance = more similar to each other ● Stratified sampling could reduce variance = strata internally homogenous (however, different from each other)