week 1 Experimental and non experimental designs Flashcards
approaches to research
- Experimental
- quasi-experimental
- non-experimental
- correlational
- descriptive
internal validity
Internal validity of a study is degree to which the study accurately answers the questions it was designed to answer. Threats to this raise doubts re reasearch limits or result interpretation. Threats include anything which gives an alternative explanation for results. Threats include; environmental variations, assignment bias, history,maturation,instrumentation differences, testing effects, regression towards the mean.
There is usually a trade off where if increase internal validity, decrease external validity.
open science and pre-registration
Doing “open science” entails hypothesis, data, methods, analysis descriptions etc are all freely available for others to view. Pre-registration is logging a hypothesis and research plan, in advance of data collection.
how to control for extraneous variables
holding constant, balance/match participants,random assignment
between-subject designs
typically a control and treatment group
within-subject designs
typically all participants are treated and repeatedly tested
correlational design
non experimental. Primarily examines whether there are relationships between variables and then seeks to describe the relationship by direction, strength and linearity. Often 2 or more measurements (one for each variable), and often scores for each variable obtained from same individual.
Pearson’s correlation coefficient measure most common linear correlation used in behavioural sciences and is -1 to 1. Spearman’s correlation is most commonly used to describe a monotonic non linear relationship(one variable tends to move consistently in a direction in response to other, but not necessarily at a consistent rate, therefore might graph more as a curve).
If both variables are none numerical (eg gender and success/failure), data codings of 0 and 1 can be used. Then obtained Pearson correlation is then a phi-coefficient commenting on the strength of the relationship but it’s direction (+/-) is not meaningful.
Note that for a study re self esteem and academic performance, a correlational study would get both scores for each individual, then assess, whereas a differential design would create low self esteem group and a high esteem group then check academic scores and compare groups.
Correlation can be used to make predictions, but does not comment on cause and effect.
Remember that strength of correlation DOES NOT equal significance. (a sample of 2 will give a perfect correlation yet unlikely to be statistically significant and a small correlation value in a large sample may be very statistically significant).
quasi-experimental designs
Designed to approximate an experiment. Typically conducted in an applied setting. Unlike a non-experimental research design, makes some attempt to minimize threats of internal validity. Aim to show a causal relationship between 2 variables but contains a confounding variable or other threat to internal integrity, which is integral and cannot be eliminated. Often groups defined due to a certain characteristic.
developmental research designs
Variation of non experimental research. eg to study changes in behaviour/ attitudes etc as age.
3 types; cross sectional, longitudinal and longitudinal-sequential.
holding constant
strategy of keeping other factors across groups as constant as possible eg same room, same instructor etc etc
balance or match participants
intentional balance group factors eg number of males and females etc
random assignment
by randomising which group a participant will be in , it is hoped that given a large enough sample, that individual characteristics will balance out across the groups.
non experimental research
the researcher cannot control, manipulate or alter the predictor variable or subjects.
Non-equivalent group design
Type of quasi experimental design. Researcher cannot control allocation to group (therefore groups may be non-equivalent)ie groups already exist. eg 2 grade three clases at different schools etc. ie Threatened by assignment bias.
3 types: Differential, post-test only, and pre-test post-test.
differential research design
Goal is to establish differences between pre-existing groups. Unlike in between-subjects research, where individual differences may be problematic, here it is the main interest.eg how is behaviour influenced by genetic differences/ how does performance differ with age?
ex post facto-looks at differences after the fact.
Correlational data and differential data give different statistical analyses but both comment on relationship WITHOUT giving an explanation or cause and effect rationale.
