evidence based Flashcards
evidence based treatment
refers to the interventions or techniques (cognitive therapy for depression etc) that have produced therapeutic change in controlled trials.
evidence based practice
broader term and refers to clinical practice that is informed by evidence about interventions, clinical expertise and patient needs, values and preferences and their integration in decision making about individual care
t-test
differente between sample means memory between walkers and sitters.
assumptions t test
variance in memory does not differ between groups (homoscedasticity), participants are independent of each other (not siblings, same housing group), memory (dep. var.) is normally distributed within groups
general linear model y(i)=b0+b1*x(i)
y= dependent variable b0= indep. var. 1 b1= different dep. var. 1 en 2
assumptions regression analysis
variances equal for all values of x, participants were independent of each other, residuals normally distributed
anova, assumptions
= can be used to test between and within subjects variables
1) variances equal in all groups
2) participants independent of each other
3) data normally distributed within groups
ancova, 2 main goals
1) correct for non-random allocation
2) reduce variance within groups (by taking into acc co-variance with a third variable), promotes power to reject H0 if H0 is really false
ancova, assumpties
1) variances equal in all groups
2) participants independent of each other
3) data normally distributed within groups
4) parallel regression lines (effect of SES on memory equal for sitters and walkers)
repeated measurements analysis assumptions
1) variances equal in all groups
2) participants independent of each other
3) data normally distributed within groups
4) sphericity (univariate approach only) (the variances of the differences between all possible pairs of within-subject conditions are equal)
multiple comparison problem
for type 1 error, if you apply 100 tests together they stand a much higher chance of false positives
power
correcte rejecting the null hypothesis (1-Beta) / , researchers consider a study to be adequately powered if it has at least an 80% chance of detecting a clini- cally significant effect when one ex- ists.
beta
type 2 error, false negative,
effect size
Effect size is a quantitative measure of the magnitude of the experimental effec// the degree of non-overlap between sample distributions (the less overlap the larger the effect size), the probability that one could guess which group a person came from, based only on their test score (effect size d=0 -> correct guess 0.5, d=1 prob rises to acceptable levels)
effect sizes for discrete variables
Cohens d, hedges g, Pearson correlation r
odds ratio
an effect size for discrete outcomes
meta regression
to get a summary of the literature
disadvantages narrative review
focus on p-values in original studies
tempration t write things that still support your theory
how to deal with studies that differ in reliability?
often based only on published literature (file drawer)
interpretation of intercept
the value. you expect if you score zero on all independent variables
r
strengst of association between variables
d
magnitude of the difference between treatment and comparison groups
NNT(needed to treat)
the number of patients who must be treated to generate one more success or one less failure than would have resulted had all persons been given the comparisons treatment
AUC (area under the curve)
represents the probability that a randomly selected subject in the treatment group has a better result than one in the comparison group
statistical power
is the conditional probability that a true effect of a precisely specified size in the population will be detected in a study using such conventional significance testing
underpowered studies
- provide insufficiently meaningful information - add to replicability problems (higher prob of false positives/negatives - are a waste of (vulnerable) clients (& everyone else’s) time
overpowered studies
- unnecessarily exposing participants to an intervention that may not work or may have side effects - are a waste of resources - (although we really want high power!)
moderation
moderators affect the effect of treatment// example : the strength of the relationship between game playing and aggression is affected by callous unemotional traits (moderator) // treatment effect depends on variables that are themselves independent of treatment (sex; nominal moderator, intelligence; continuous moderator)
mediation
mediation is said to have occurred if the strength of the relationship between the predictor and outcome is reduced by including the mediator// mediators are affected by treatment//which processes are important during an intervention? (what mechanisms underlie intervention effects) (sleep problems; nominal mediator, hours of sleep ; continuos mediator)
logistic regression
regressing a dichotomous dependent variable on the basis of on or more (nominal or continuous) independent variables
with logistic regression you can do enter or stepwise
enter = keep all independent variables in model (test an explicit hypothesis : confirmatory) or stepwise forward/backward = let sass find out which variables are important (no explicit hypothesis : exploratory)
evidence based clinical practice
clinical practice based on the (best) available evidence combined with the practitioner’s personal expertise and experience, adapted to the patient’s needs and preferences
intention to treat analysis
based on the original treatment assignment and not on the treatment eventually received /completed. intended to avoid potentially misleading artefacts that can arise in intervention research, such as nonrandom attrition of participants
per-protocol analysis
a comparison of treatment groups that includes only those patients who completed the allocated treatment -> may lead to bias: depending on the distribution of dropouts, randomisation may no longer be OK. used to see if
relative risk
the odds that improvements are greater in the control condition than in the intervention condition •RR = odds of impr. in control condition / odds of impr. in intervention condition–we want this ratio to be 1 or lower
absolute risk reduction
absolute difference in the odds of improvement in the control condition and in the intervention condition •ARR = odds of impr. in intervention condition - odds of impr. in control condition–we want this to be 0 or lower
relative risk reduction
reduction in the incidence of negative outcome•RRR = 1 – RR–we want this to be as close to 1 as possible
multilevel modeling
designed to analyse data that exist at different levels / analysis of data gathered in multiple groups (intervention study in multiple nursing homes)//dependent variable : continuous, independent variable : nominal or continuous
all variables : manifest(are observable)
unit of analysis problem
when the moderator variable exists at a different level than the independent and dependent variables, should the analysis be done at the level of the student or at the level of the program ?
level 1 MLM independent varibale
client characteristic
level 2 independent variable
therapist characteristic
individual patient meta regression
multiple patients from multiple studies. can incorporate : level 1 independent variables (related to individuals) and level 2 independent variables (related to studies)
2 types of theries
manifest variables, latent variables, relations between manifest variables (networks)
disadvantages of nominal latent variable based on concensus
high comorbidity, arbitrary limits (why 6 out of 9?), heterogeneity within groups : extreme cases 6 out of 12 symptoms (one diagnostic category may contain two completely different people)
mixture analysis model
nominal
factor analysis
continuous
factor mixture analysis
both nominal and continuous
advantages of nominal model
no comorbidity, no arbitrary limits, reduced heterogeneity within groups
network theory
“this person is distracted, and therefore has trouble organising tasks” focused on the relationship between manifest variables = symptoms
purpose of mixture and cluster analysis
determine latent groups (within groups equal and between groups different
differences mixture and cluster analysis
mixture is a statistical method. cluster is not a statistical method (cannot test hypotheses, is exploratory, no assumptions)
steps in cluster analysis
selection of sample (theory) selection of variables (theory calculate inequalities between participants do the cluster analysis interpretation of cluster analysis choosing number of clusters validation of final cluster solution
chi square test
relationship between two categorical variables. statistic is based on the idea of comparing the frequencies you observe in certain categories to the frequencies you might expect to get in those categories by chance
fishers exact test
a solution to the fact that if the cell is <5 the test statistic is too deviant from a chi square distribution to be accurate. the fishers exact test is a way to compute the exact probability of the chi square test in small samples
likelihood ratio
alternative to Pearson chi square, based on maximum likelihood theory. general idea is that you collect some data and create a model for which the probability of obtaining the observed set of data is maximised, then you compare this model to the probability of obtaining those data under the null hypothesis
loglinear analysis
analyse more complex contingency tables in which here are three or more variables
assumptions when analysing categorical data
independence of residuals (each person item or entity must contribute to only one cell of the contingency tabel) and expected frequencies (no expected values should be below 5
centring variables
‘use of grand mean centring’, refers to the process of transforming a variable into deviations around a fixed point. this is used when it makes no sense that the b in the equation would be zero (like heart rate)
simple slopes analysis
comparing the relationship between the predictor (time spent gaming) and outcome (agression) at low and high levels of the moderator (callous traits). the essence is that we work out the model equations for the predictor and out come at low high and average levels of the moderator
mediation direct effect
pornography consumption on infidelity, the relationship between them controlling for relationship commitment (c’)
mediation indirect effect
the effect of pornography consumption on infidelity through relationhship commitment (a-b+c’)
network structure
the web of relations among symptos
network state
the activation of symptoms
causality hypothesis
when causal relations among symptoms are strong, the onset o tone symptom will lead to the onset of others
connectivity hypothesis
strongly inter connected symptom networks are thus vulnerable to a contagion effect of spreading activation through the network
centrality hypothesis
higher central symptoms (stronger inter symptom connection) have greater potential to spread symptom activation throughtou the network than do symptoms on the periphery
comorbidity hypothesis
because some symptoms occur in multiple disorders, symptom activation can spread between syndromes, with symptoms bridging these syndromes playing a critical role in psychiatric comorbidity
level of momentary experiences
symptoms are aggregates of moment to moment experiences. these micro processes constitute the true building blocks of psychopathology
conditional positive manifold
even after controlling for shared variance among symptoms, these symptoms tend to be positively interconnected
what is de dependent variable in meta regression analysis?
effect size
when you have an intercept (so no independent variable) in meta regression analysis
the interpretation of the intercept will be the overall effect size across the difference in side effects between placebo and drug users
N=1 analysis
is comparing the baseline before the intervention with the period of during the intervention
AR1
autoregressive of order 1. correlation timepoints is only dependent on their distance ‘lag’
normative comparisons
comparing the score of an individual to that of a typical sample. essence of psychological assessment
family wise error
chances of at least one a-typical test result , problem : FWE increases with number of tests
interpretation of gamma zero zero parameter
it is the score on a test (dependent variable) in a given class if the fear of failure level of the level 1 independent variable is zero
