why is random sampling often not possible for the vast majority of research?
not possible due to fixed conditions, not practical, not always desirable
probability sample: likelihood that any individual in the population being selected can be specified, most common type is simple random sampling
non-probability sample: likelihood that any individual in the population being selected can not be specified
when are random samples most desirable?
when the researcher wants to accurately describe a population
probability samples, including random samples, are virtually never used in psychological research. Psych research does not describe how a population behaves, it tests the relationship between psychological variables
A representative sample, as opposed to a random sample, is always desirable
sampling error: differences arise between sample results and the parallel results when the entires population is measured
the sampling error can be estimated with the error of estimation, margin of error and confidence intervals
what factors influence the margin of error in a probability sample?
sample size, population size, variance of the data
winning 45% of a vote with a 3% margin of error means 95% of the time, between 42% and 48% of the vote is won
simple random sample: every member of population has equal probability of being selected. in small samples, can be done mechanically. in large samples, computerized randomization is used
systematic sampling: select some starting point and the select every nth element in the population (e.g. "every fourth person"
stratified random sampling: divide population into strata (gender, ethnicity, etc.), select randomly from each stratum. political polls
which methods require a sampling frame (where all members of pop. are accessible)?
simple random sampling and stratified random sampling
cluster of multi-stage sampling: for very large populations where everyone cannot be accessed
misgeneralization: generalizing to a population not represented in sample
nonprobability sample: any sample in which little or no attempt is made to get a representative cross section of the population. convenience sampling, quota samples, purposive samples
convenience sample: most common type in psych research. usually college students, often adequate, use different samples for comparison
quota samples: convenience sample where certain types of participants are obtained in particular proportions (e.g. ethnic breakdowns)
purposive sampling: researchers use past research findings or judgement to decide who to include in sample (election sampling)
participantnumberdilemma: a large number of subjects will provide more valid results, but we must weight the cost/benefit of acquiring more subjects
type 1 error: we say there is a difference between groups or a variable relation, but there is not
type 2 error: we say there is no difference of variable relation, but there is
p-value: significance
power: probability of finding a significance between groups/variable relation
a higher alpha gives more power. usually 0.5 higher alpha means a higher chance of a type 1 error
one-tailed tests are more powerful than two-tailed tests. directional hypotheses more valuable
a larger effect size gives more power
a larger sample gives more power than smaller samples. resources limit this
correlation: relationship between two or more variables
how do we determine whether one variable is affected by another?
we see if the scores of these variables covary (vary or change together)
bivariate correlation: simple correlation between two variables
correlations are designed by the letter r, which ranges between -1 to +1
a correlation is also known as the personproduct-moment or the personr
the pearsonr is predicated on the assumption that the two variables involved are approximately normally distributed
when variables are not normally distributed, the Spearman correlation is more appropriate
perfect correlations are almost never found in the social sciences, primarily a mathematical or theoretical construct