Multivariate Analyses Flashcards by Will Perchard

When is multivariable analysis used?

For data with one dependent variable but more than one independent variable

How well did you know this?

Not at all

Perfectly

What is multivariable analysis used to determine

Relative contributions of different causes to a single event or outcome

How well did you know this?

Not at all

Perfectly

When is multivariate analysis used?

For data with more than one dependent and independent variable

How well did you know this?

Not at all

Perfectly

What is used if both dependent and independent variables are continuous?

Multiple regression

How well did you know this?

Not at all

Perfectly

What is used if the dependent variable consists of dichotomous categoric data (two outcomes)?

Logistic regression

How well did you know this?

Not at all

Perfectly

What is used if the dependent variable includes a time factor e.g. a survival curve?

Cox proportional hazards model

How well did you know this?

Not at all

Perfectly

What is used if the dependent variable consists of nominal categorical data i.e. more than 2 outcomes

Log-linear analysis

How well did you know this?

Not at all

Perfectly

What is used if dependent variable is continuous and independent is categorical?

ANOVA

How well did you know this?

Not at all

Perfectly

What does ANOVA stand for?

Analysis of variance

How well did you know this?

Not at all

Perfectly

What is used if there are both categorical and continuous independent variables?

ANOCA

How well did you know this?

Not at all

Perfectly

What does ANCOVA stand for?

Analysis of covariance

How well did you know this?

Not at all

Perfectly

What is path analysis?

Extension of multiple regression

How well did you know this?

Not at all

Perfectly

What can path analysis do?

Examine situations where there are several final dependent variables with chains of influence

How well did you know this?

Not at all

Perfectly

What is cluster analysis?

Multivariate tool used to organise variables into homogenous groups or clusters

How well did you know this?

Not at all

Perfectly

What does cluster analysis involve the generation of?

A similarity matrix

How well did you know this?

Not at all

Perfectly

What does a cluster analysis produce?

A dendogram

How well did you know this?

Not at all

Perfectly

What is canonical correlation?

A multivariate tool used to explore the relationship between two sets of variables

How well did you know this?

Not at all

Perfectly

What does canonical correlation involve?

Computation of eigenvalues

How well did you know this?

Not at all

Perfectly

What is discriminant function analysis?

Multivariate technique used to detect which of several variables best discriminates between 2 or more groups

How well did you know this?

Not at all

Perfectly

What is discriminant function analysis similar to?

Computationally similar to MANOVA

How well did you know this?

Not at all

Perfectly

What does MANOVA stand for?

Multivariate analysis of variance

How well did you know this?

Not at all

Perfectly

Key uses of multivariate analysis

Look for interaction between independent variables
Quantify associations
Adjust for potential confounders in controlled study
Develop models to predict values or probabilities of certain outcomes

How well did you know this?

Not at all

Perfectly

What test to use if dependent variable is nominal categorical?

Log linear analysis

How well did you know this?

Not at all

Perfectly

What test to use if dependent variable is categorical dichotomous?

Logistic regression

How well did you know this?

Not at all

Perfectly

What test to use if dependent variable is categorical dichotomous with a time factor?

Cox proportional hazards

What test to use if dependent variable is continuous and independent variable is categorical?

ANOVA

What test to use if dependent variable is continuous and independent variable is continuous?

Multiple regression

What test to use if dependent variable is continuous and independent variable is categorical and continuous?

ANCOVA

What does factor analysis refer to?

Set of statistical methods used to detect underlying patterns in relationships among a number of observed variables

Aim of factor analysis

To identify whether correlations between a set of multiple observed variables can be summarised in terms of a smaller number of underlying, latent, unobserved variables called factors

Two approaches of factor analysis

Exploratory Confirmatory

When is exploratory factor analysis used?

For preliminary investigation of a set of multiple observed variables

What does exploratory factor analysis not do?

It does not make an a priori assumption about the composition of underlying latent variables or factors

Applications of exploratory factor analysis

Data reduction when multiple (>25) variables have been measured, to provide a parsimonious description of the data. Classification of sx into clinically meaningful concepts Definition of subscales of new measures

What is confirmatory factor analysis used for?

To test whether a specified factor structure remains valid with new dataset

When is confirmatory factor analysis used?

For assessing construct validity of questionnaires or tests

Stages of factor analysis

Construction of correlation matrix Extraction Rotation measuring the eigenvalues Define factors to be retained Include as many factors as required Labelling

What types of methods are used in extraction?

Common factor analysis Principal components analysis

What is the eigenvalue

Eigenvalue is the amount of total variance explained by each factor

How can we determine the number of factors to be retained?

Kaiser rule Scree plot Include as many factors as required until adequate preset proportion of variation is explained by them

What is kaiser rule?

Only factors with eigenvalues greater than 1 are retained

What is scree plot?

Plot the component numbers against eigenvalues. Choose the number that forms the bend before the plot levels off on the right side

What happens during labelling and reporting factors?

There is general consensus that variables with a factor loading grater than or equal to 0.4 are probably making a significant contribution to that factor in contrast to those with smaller factor loading

What is path analysis?

Causal modelling and prediction beyond simple regression

What happens in path analysis?

A set of independent variables is related to a set of dependent variables

When is path analysis used?

To examine situations where there are various chains of influence amongst variables studied.

How is a path relationship displayed?

Using arrows that display presumed causal relations.

What is the name of the independent variable in path analysis?

Exogenous variable

What is the name of the dependent variable in path analysis?

Endogenous variable

What does single headed arrow mean in path analysis?

Flows from putative cause to effect

What does double headed arrow suggest in path analysis?

Mere correlation but no predictive, causal link

What is a path coefficent in path analysis?

Indicates the direct effect of a variable assumed to be a cause on another variable assumed to be the effect

What are the two subscripts with which path coefficients are written?

P21 is the path from 1 to 2 i.e. effect is listed first

What is recursive in path analysis?

Path analysis in which the causal flow is unidirectional

When is stratification used?

To control for or analyse the effect of confounder variables

What is a stratum?

Sub-group within a sample often defined by presence or absence of variable of interest

Importance of stratifying data according to confounder variables

Later one can analyse each strata to find the degree of association between presumed cause and effect

What is adjustment in stratification?

When we later produce a single overall estimate using various methods for obtaining summary risk values

Why is a method of weighting suggested for stratification?

Crude summary may not reflect actual risk

How can one use weighting for stratification?

Mantel-Haenszel procedure

What can Mantel-Haenszel procedure be used for in stratification?

Weighting Help to find if variable is just an efect modifier or real confounder

What is standardised?

Method used in large data sets for public health statistics to produce adjusted rates

What is used for standardisation in public health?

Hypothetical standard population produced by WHO

Types of standardisation

Direct Indirect

What happens in direct standardisation?

Stratum specific rates from study sample are applied to standard population

What does direct standardisation produce?

Summary score

What happens in indirect standardisation?

Stratum specific rates from standard population are applied to study sample

What does indirect standardisation give?

Expected rates

How does one arrive at standardised rates from expected rate?

Expected rate is divided by the observed rate

What type of confounder is stratification useful for?

Known confounders only

What can adjustment be applied to?

Relative Risk Odds ratio

What happens to sub-strata risk if the third variable is neither an effect modifier nor a confounder?

Sub-strata risk does not differ from crude total risk

What happens to adjusted vs non-adjusted risk if the third variable is neither an effect modifier nor a confounder

Summary risk does not differ much from crude total risk

What happens to sub-strata risk if the third variable is both an effect modifier and a confounder?

Sub-strata risk differs from crude total risk

What happens to summary risk if the third variable is borth an effect modifier and a confounder?

Summary risk differs from crude total risk

What happens to sub-strata risk if the third variable is an effect modifier but not a confounder?

Sub-strata risk differs from crude total risk

What happens to summary risk if the third variable is an effect modifier but not a confounder?

Summary risk does not differ much from crude total risk

What happens to substrata risk if the third variable is a cofounder but not an effect modifier?

Substrata risk does not differ from crude total risk

What happens to summary risk if the third variable is a confounder but not an effect modifier?

Summary risk differs from crude total risk