Module 7 - Lecture notes 7 PDF

Preview

Summary

katie gunnel...

Description

Module 7 7.1 WHo do we measure? - When do we care about generalization? - Interrogating external validity - External validity: how representative results generalize to pops not in our study - Two types - External validity of sample: - Ppl - Does sample generalize to pop of interest - External validity of setting - Contexts - Do ratings of prof for one class generalize to same prof in other class? -

Population: entire set of individuals of interest to researcher - Ex. psych 2001 A students - Use samples - Subset of population Researchers are interested in specific populations - Population of intertest: - More limited - Ex. elderly ppl with dementia in saskatchewan Sampling: variety of ways selecting individuals from pop of interest - To be generalizable, sample must come from the pop - Doesn't mean it does generalize to every person in pop of interest Biased sampling (unrepresentative sample) - Some members of pop of interest have much higher probability of being included compared to others Unbiased sampling - All members have equal chance - Only unbiased is generalizable Sample representativeness: extent  to which characteristic of sample reflect those of population - Ex. ppl in our sample represent demographics like ppl on campus and off

7.2 Probability sampling - Involves recruitment techniques that have some level of randomness

-

Every member of pop of interest has equal and known chance being selected Random samples of good external validity - List of participants must be available - Know proportions of ppl and how likely are to get into sample - Use unbiased method of selecting: random - Every possible outcome equally likely Probability sampling techniques -

-

-

-

Simple random sampling: - Mot basic - Ex. list of ppl in pop of interest. Assign them number and use random number generator - Advantages: - Completely random - Effective practical of representative sample - Weakness: - Impractical if population is large - we wont necessarily have a list of pll to put into generator - All canadians would not work - Expensive if population is widely dispersed - Chance determines each selection therefore its possible to obtain distorted sample ex. Randomly get only females Cluster sampling: - Ppl are already divided into groups - Ex. high schools in ottawa, each highschool is a cluster then randomly pick cluster rather than individual ppl Multi-stage sampling: - Two random samples selected 1. Random sample of clusters 2. Random sample of ppl within clusters - Ex. take highschools in ottawa - randomly select high school then ppl in it - or teams Stratified random sampling: - Researcher purposefully selects particular demographic categories or strata then randomly select individuals within - Representative of ethnic diversity in canada - Want subgroups to be adequately representative - Ex. - 1. Identify subgroups (strata) based on ex (age, income) - 2. Determine proportion of population in each subgroup - Perhaps with statistics canada

-

-

-

-

3. Select sample so proportion matches proportion of overall population using a random sampling technique Advantages: - Guarantees sample will contain subgroups of pop that that match pop of interest that match proportion Weakness - Not always easy as might not have enough info to make strata

Oversampling: - Researcher intentionally over represents one or more groups - Variation of stratified random sampling - Ex. sample of 1000 ppl but you wanna make sure south asian ppl are well represented, you might oversample that population - but still random Systematic sampling: - Using a random numbers generator, the researcher selects a number at random and then counts ppl in the room. - Ex. every 7th person included

7.3 Non-probability sampling \ -

-

How do researchers end up with biased samples? - 1. Convenience - Sampling easy to contact - 2. Self selection - Sampling only those who invite themselves - Not random now Non-probability sampling: non-representative  sam[ling method - Odds of selecting someone into sample is not known sas don't know enough about population of interest - Wouldn't have info of whose in pop - Dont use random sampling so not representative - Convenience sampling: using sample of ppl who are easy to contact - Common in psych - Collect data that are easy to get accessibility - Ex. sona - WEIRD science - What we know in psych is mostly from undergraduate students which does not generalize well - Using MTurk (amazon participants) - Ppl making money for completing studies

-

Recruiting from population - Researchers might approach ppl differently - Ex. putting up posters - Advantages of convenient sampling: - Easier - Less expensive - Less time Disadvantages - Weaker form of sampling - use posters in grocery store - only reaching to ppl who go into grocery store - No attempt to use random selection - Little control over representatives of sample - likely biased - Self-selection sampling: - When a sample is known to contain only people who volunteer to participate - Common among online surveys - Causes severe threats to external validity - Because we do not know thee degree of difference between those who self select vs. not but speculate they might have stronger opinions - Rate my prof ex. More likely to have strong opinion versus school rating which would be more representative Four different types of non probability sampling 1. Convenience sampling a. Purpose is used when you want to study certain kinds of people, so you only recruit them. Ex. ppl with osteoporosis 2. Snowball sampling: a. Have participants, then say if you know anyone else with certain thing or part of certain group share with them b. Variation on purposive sampling c. Not representative as not random 3. Quota sampling a. The researcher identifies subsets of pop and set target number of ppl from each subset. Researcher uses non random sampling to find ppl like convenient sampling 7.4 External validity When is external validity highest priority? - In frequency claim

-

Frequency claims are about how often something is occurring. Want to ask external validity because if they use non probability it could distort frequency When is external validity a lower priority? - Researcher looking at associations or casual links - they can still be accurately detected without representative samples - Ex. conscientious person reports traffic in area - that person is different to average as clearly more conscientious, but when looking at traffic, that will be there for every person regardless of if they are conscientious Interrogating external validity: - Larger sample sizes are not more representative - Sample size has to do with statistical analysis - Need to look at how participants were recruited - Ex. canadian election wants to run frequency claims on party votes. If polled in conservative convention - not representative - even if 5000 ppl - Ppl think results will generalize to pop if lots of participants - that's more with statistical validity then external - Statistical validity looks at range and margin of error - which is lower with more participants - But external validity looks at generalization Random sampling is creating a sample with random method so that each member of pop of interest has eque=al chance - Increases external validity Random assignment is once ppl are in study - how do you split them up into diff conditions and groups within experiment - Increases internal validity...