Types of Sampling

Inference: Drawing conclusion about the population on the sample data Sampling variability: Variation in the value of.a statistic from sample to sample. This could be affected by selection bias and sampling error Selection bias: systematic favouritism in data selection process. Sampling error: difference between the sample and the parameter. This could happen by sample not covering the whole population, where some groups of individuals with certain characteristics are left out. Law of Large number: If we keep taking larger number of individuals, we could get closer the parameter(to the characteristic of the population) Sampling with replacement: Individuals are returned to the population after data collection, to potentially be selected and measured again Non-probability sampling: impartial chance is not used when selecting units. Tend to be selected from how easy they could access the sample Convenience sample: choose individuals by who is easiest to find Voluntary response sample : Individuals are self selected in response to open call/invitation. Emails, phone survey. Bias because only people with certain strong opinions are likely to respond. Non-response bias: Some individuals do not respond to the survey. Types of Sampling Probability sampling: Involves selecting a unit using impartial chance. All individuals would have equal chance of selection Random sampling: All unit have equal chance of selecting Simple Random Sampling: each uni of the population has an equal chance of being selected. All possible combination of size n from the population are equally likely Stratified Sampling: sampling frame is subdivided into 2+ groups based on variables by characteristics. Samples are retrieved from these subsets. Must know the characteristics of the whole sampling frame. Good if there's a group of minorities. Cluster sampling: population divided into 2+ clusters based on selected variables then a subset of the cluster are censused Systematic Sampling: unit is selected at the starting point and then additional units are sampled at a constant interval k. Good when the order does not matter, alphabetical names for example Multi stage sampling: Samples are selected from at least two iterations of selection of individuals from the sampling frame. Ttpes of studies ●​ Observationaly study: obserers individuals and measures both explanatory and resonse variables but does not attmeps to impose conditions on the units or influence the response variable but does not attempt to impsoe conditoon on the unit or influenc response variable. ○​ Could have cofoundning varable which is a additional variables which effect the response variable ○​ Cohort study ○​ Survey ○​ Case control observational study ●​ Experimental study: researchers give out the condion or alter the explanatory varabe to retrive data. ○​ Full factorial ○​ Control Types of experimental studies ●​ Completely randomized design: samples are randomly assigned to the treatment Ex) A group of student is each randoomly given treatment and the data was collected after the treatment ●​ Randomized block design: Sample is divided into block on pre-exisiting characteristic(similar to stratified sampling) ○​ Treatment gourps at the largest scale have similar composition ○​ The desing allows resercher to evaluate wheter an interaction exists between the treament effects and the blocking variables ●​ Matched pair design: pairs of units from the sampel are mathed base on similarity across pre existing characteristics, and then, randomly assing the pairs to treatment ●​ Repeated measurement desing: each unit in the sampli is assigned to both or all treatment ○​ Response should be calculated in the same order ○​ Order of treatment is random to make sure on does not affect the other. Measurements ●​ Measurement error: deviation caused by factors that randomly affect measurement across the sample ●​ Measurement bias: deviation caused by factors that systematically affect measurement across the sample ●​ Accuracy:measurement approaches the true value on average(opposite:bias) ●​ Reliability: measurement that produce the smap value for repeated application on the same individuals ○​ Reliable measurements have little to no measurement error ○​ Quantifying degree of variation between measurements that can gauge reliability ●​ Validity: measurement that actually characterize the variable of interest ○​ Using different types of measurement to characterize a variable will allow researchers to determine if there is consistency amongst the measurement of samples. Dealing with explanatory variables ●​ REducing variation: accounting for variation in other variables to isolate the effect of the factors of interest. There are two ways of doing so ○​ Hold other variables constant ○​ Seelect a narrower sampling frame ●​ Important variable into the study design ○​ Blocking: subdivision of sample units based on pre-existing variables prior to assignment to treatments ○​ Blinding: condition in which the research or study participants are unaware of the treatment to which the units belon ■​ Placebo: treatment that is fake but otherwise indistinguishable from experimental groups ○​ Comparisons:incorporate in more than one treatment into a study to allow comparisons for natural change. ●​ Exploration: using chase to assign individuals/units from the sample tor treatments or the treatment order ●​ Consequences: randomization created treatment groups that are similarly composed on average. ●​ Replication: inclusions of more than one individuals from the same individuals instead of many individual ○​ Pseudo-replication: several measurement from the same individual instead of many individual( all measurements are still considered 1 replication)​ ■​ May help determining how reliable four measurement process is but they don't capture the variation in the response variable that exist between individuals