research methods and longitudinal research

Internal validity

Accuracy and truthfulness of study results

• how well was the study conducted

• Could the results be due to other factors?

External validity

Generalizability of the study results

• how applicable are the findings to other people and settings in the real world?

Measures

• validity

• Reliability

Validity

Did they measure what they intended to measure?

• operational definition

• Ex. How do we measure happiness? Or aggression?

Reliability

Consistency of results over time

• A measure can be reliable and not be valid, but a valid measure must be reliable

Cross sectional study

A study that compares two or more groups or individuals at one point in time.

• benefits: study done quicker because it’s done at one point in time, and cheaper

• Limits: cohort effects (a group of people who share a common experience and historical experience)

longitudinal study

A study that follows the same group of individuals over time.

Developmental research designs

• cross sectional study

• Longitudinal study

• Sequential research design

Sequential research design

Includes elements of both longitudinal and cross sectional designs

Longitudinal advantages

• examines changes within individuals o er time

• Allows for developmental analysis

Cross sectional advantages

• examines differences between participants of different ages at the same point in time

• Efficient and cost effective

Sequential advantages

• examines changes within individuals over time

• Examines differences between participants of different ages at the same point in time

• Can be used to examine cohort effects

• Can be used to examine time in history effects

Sequential disadvantages

• expensive

• Time consuming

• possibility of practice effects

• Some participant attrition

Cross sectional disadvantages

Longitudinal disadvantages

Selection bias

How the participants were selected for the study leads to them being different from the target population or not representative

• volunteer participants

• Convenience sampling (where researchers are recruiting samples from)

Merits of Longitudinal Research Grammer et al. (2013)

• exploring developmental trajectories

• understanding the origins of later abilities

• Relating Interacting processes over time

• The importance of context as a mechanism for change

• Repeated assessments

• Using existing large data-sets

Exploring development trajectories

Developmental trajectories of skill acquisition allowing examination of inter individual (between individuals) and intra individuals(within an individual) variability

Understanding the origins of later abilities

Relating Interacting processes over time

• allows for examination of.p mechanisms that may underlie development change

• Exploration of associations between related skills over time

Relating Interacting processes over time

• allows for examination of.p mechanisms that may underlie development change

• Exploration of associations between related skills over time

Repeated assessments

• practice effects or behaviour modified as a result of repeated assessment

• Do the measures operate similarly at different ages?

• What type of Controls are needed?

• Issues related to homotypic and heterotypic continuity

• Including too many measures in the proposal

• Longitudinal research is still correlational - can not determine causality

Homotypic continuity

One disorder or trait predicts the same disorder later on (ex. Early anxiety predicts later anxiety)

Heterotypic continuity

One disorder predicts a different disorder later on (ex. Early anxiety associated with depression later on)

Using existing large data-sets

• early childhood longitudinal program (ECLS), NICHD dataset - data already collected

• Many population based studies may be used to answer questions they were not originally designed for

• Large samples with power to examine complex associations over time

• Measures available often differ from those that would be included in a newly developed study

Power

Ability to detect a difference if there is a difference

• influence by sample size, 80% power considerable acceptable

Type 1 error

Conclude there is a difference between groups when there is NO difference

• false positive

• Alpha level = 0.05% chance of making a type 1 error (vs 0.01 = 1% chance)

Type 2 error

Conclude there is NO difference when there IS a difference

• false negative

• 80% power means there is a 20% chance you will miss finding an association

Collecting longitudinal data

• choose the best research design

• Select target age-range of participants and how long to follow them

• How often to assess them and on what measures

Strategies for conducting longitudinal research

• building and maintaining good relationships with communities of research participants

• Working to develop a skilled staff of researchers

• Conducting child centred research

• Giving back to the individuals who have participated in the research

  • for children, it’s given in groups, not individually

CHILD study

• largest multidisciplinary, longitudinal birth cohort study in Canada

• Recruited over 3500 pregnant women who gave birth between 2009 and 2012 in Vancouver, Edmonton, Winnipeg, and Toronto

• Children have been followed since birth to 13-16 years old

• Assessed on a variety of different outcomes at different ages

• Allergies, asthma, pollution, diet, pregnancy factors, home environment, neurodevelopment, sleep, screen time, physical activity

Confounding variable

A third variable that distorts the relationship between two variables

• creates a misleading association

Control for confounding variable

• restriction

• matching

• Stratification

Restriction

Exclude some participants

• limits sample size and power [the ability to detect a difference if there is a difference]

Matching

match participants in each group on a variable (e.g. age, SES)

Stratification

Create subgroups (strata) on different levels of confounding variable

• number of participants in each stratum is reduced - reduces power

Multiple regression

• multivariate technique

• Association of multiple independent variables on one outcome variable (dependent variable)

• Examine the unique effect …