ch 7: survey research

self-report

respondents report their own thoughts, feelings and behaviour, by questionnaire or interview

careful sampling

the goal is an accurate estimate of what is true in a population, so large random samples are preferred

representative sample

resembles the population in evert way that matters for the question

sampling frame

list of the population that you can sample from

non-probability sampling

you cannot state the chance of selection

• convenience: nearby and willing

• snowball: participant recruit others

• self-selection: people volunteer on their own

probability sampling

you can state each person’s chance of being selected

• simple random: everyone has equal chance

• stratified random: split list into groups (strata), then sample at random within each

• cluster: randomly select groups, then people within them

sampling bias

how the sample is chosen makes it unrepresentative, so results are inaccurate

non-response bias

the people who do not respond differ from those who do

a perfectly random sample can still have non-response bias

sampling error

gap between a samples result and teh populations true value, caused by who happened to be drawn

• random not systematic

frequency table

how many chose each category, report percentages

bar graph

order by frequency, largest first, seperate categories - not continuous variables

statistically significant

a difference this large would be unlikelt if there were no difference in the population

what is p

the probability of a difference at least this large, if there were really no difference in the population

five steps to one answer

1. interpret

2. retrieve

3. judge

4. format

5. edit

why self-reports go wrong

people cannot tell you

• why

• memory

• future

people will not tell you

• social desirability: looking good

• demand characteristics: answer researcher seems to want

• response sets: agreeing with everything

open-ended

respondents answer in their own words

use when you dont know the range of answers

slower; answers must be coded

closed-ended

respondents choose from options you give them

use when the constructs is well-defined

quicl to score, but the options steer answers

good practice in rating scales

• 5-7 options

• label options with words

• balance the options around the midpoint

bruso

• brief: not wordy

• relevant

• unambiguous: clear to understand

• specific: dont ask double barrel

• objective: neutral questions

putting the survey together

• into: why it matters, whos running it, informed consent

• clear instructions, give example

• most important items first

• easy items before sensitive ones

• demographics last

• end with thanks