introduction Flashcards
2 types of statistics
descriptive
inferential
descriptive stats
describe study of population
inferential stats
using what we know from data we’ve collected to make inferences about what we don’t
cycle of experimental research design
current state of knowledge
construct hypothesis to test
design experiment
execute experiment
carryout stats analysis
interpret and report
what’s a research design
framework / blueprint for conducting research project. details procedures to obtain correct info needed to solve research problem
importance of research design
provides smooth operation
makes research efficient
blue print for advanced planning
precautions to reduce errors
reliability depends on design
what makes a good research design
specifies sources and info needed
strategic roadmap for collecting and analysing data
defines timelines and cost
include statement of problem, procedures and techniques, range of processes and analysis
criteria of research design
reliability
replication
validity
types of validity
measurement
internal
external
ecological
what are variables
main focus of research
characteristics of variables
continuous (measure in a specific way and even intervals)
discrete (specific items)
categorical (eg mode of transport)
dichotomising continuous and discrete (convert with to categorical)
4 levels of measurement
interval scale
ratio scale
nominal scale
ordinal scale
interval scale
scores in order of magnitude with equal intervals
ratio scale
same as interval but has absolute zero point
nominal scale
attributes are only named
ordinal scale
attributes are only ordered
extraneous variables
might impact on variables we are interested in but failed to take into account
confounding variables
types of extraneous variable related to both main variable and one we are interested in
research designs - experimental
correlational
experimental
quasi
research designs - cross sectional
longitudinal
case study
correlational design
relationship / association between variables
stats techniques used for correlational design
Pearson product of moment correlation
spearman’s rho
simple linear regression
multiple linear regression
chi squared analysis
issues with correlational design
not causation (only observing and recording changes)
experimental research design
experimenter manipulates the IV and see an effect on the DV (differences between conditions)
features of experimental research
has a hypothesis
random allocation
quasi research design
seeing if there are differences on the DV between conditions of the IV (pre existing difference)
no random allocation
quasi stats techniques
t test
mann whitney
wilcoxon
analysis of variance
structure of experimental design
has a time line:
obs = observation in relation to DV (pre and post test)
exp = experimental treatment
no exp = control
t = timing of obs
within participant design
repeated measure / related
same ppt in every condition
between participant design
independent / unrelated
different groups of ppts
+ve of within participant design
control individual confounding variables
fewer ppts = low cost
-ve of within participant design
order effects - practice, boredom, fatigue
counterbalancing
demand characteristics
+ve of between participant design
less likely to get bored or tired = less order effects and demand characteristics
-ve pf between participant design
more ppts
lose some control over confounding variables
individual differences
cross sectional design
collection of data on more than one case at a single point in time with 2 or more variables
interested in variation and relationship between variables
structure of cross sectional design
collection of data on a series of variables at a single point in time
creates a rectangle of data
features of a questionnaire
reduced bias (presented to everyone the same way)
open and closed questions
descriptive research
large samples
self administered or by researcher
less intrusive
postal versions
easy to do and analyse
pre coding questionnaire
set format closed responses
can be scanned into SPSS data sheet
designing a self completion questionnaire
easy to follow
don’t cramp
identify response sets in a likert scale
clear instructions how to respond
keep questions and answers together
well worded questions
easily understood
happy to share info
start and end with ‘friendly’ questions
avoid 2 in 1
use simple language
avoid leading
don’t presuppose
ensure produce variablilty
test in person with trial
vary format to reduce boredom
include response id on every page
+ve of self completion questionnaire
cheaper
quicker
no interviewer effects
no interviewer variability
convenience for respondents
-ve of self completion
cannot prompt
cant ask too many
difficult to ask other kinds of questions
can be read as a whole
cant collect additional data
can report desirable answers not real
may not take seriously
not appropriate for all
risk of missing data
low response rate
how to improve response rate
good covering letter
return S.A.E with clear instructions
follow up for no responses
questionnaire correct length
capture interest
limit open ended questions
provide incentive eg voucher
what does a cover letter include
friendly
short
describe what’s being done
tailor to audience
incentives
s.a.e
use deadline
describe confidentiality
name and phone number
features of follow ups
include 2nd copy of questionnaire
cant guarantee anonymity only confidentially
sounds surprised at non response
how can bias occur with random allocation
due to missing data
not randomly missing data
fails to reply to question because they don’t want to answer
pilot study
tests measures too are going to use on a very small sample and helps identify the feasibility
guidelines for creating data for analysis
decide variable and document them
design data set with one subject per line
variable with proper designated name
descriptive labels for each variable
select type for each variable
additional tips for categorical variables
define missing values codes
consider need for grouping variables
variable with proper designated name
begin with letter
short
no blanks or special characters
unique and no duplicates
case doesn’t matter
descriptive labels for each variable
eg Age: Age on 1st Jan 2022
Gender: male or female
select type for each variable
numeric (quantitative)
character (categorical)
additional tips for categorial variables
case consistent
avoid long data codes
consider binary coding
what not to do with missing data
don’t fill with 0 eg use ‘’’
data screening
check errors (scores out of range)
find error in data file
correcting the error in data file