medical statistics

• what is the difference between frequentism and bayesian?

  frequentism - the frequentist approach assigns probabilities to data, not to hypotheses,

  bayesian - approach assigns probabilities to hypotheses
• what is the difference between ontology and epistemology?

  ontology -what is true/what we know about the world

  epistemology - how we can come to know it
• what do ontology and epistimology mean in quantiitative research?

  what does positivism mean and how does this relate to positivism?

  • a concrete objective reality we can come to measure
  • "scientific method"
  • positivism = scientific method, only believe what can be proven
• variables: dependent, independent, confounding/extraneous, control

  what is the role of variables?

  dependent variable - measurement

  independent variable - what is being measured

  confounding/ extraneous - affect dependent and independent variables

  control variable -has to be kept at a constant rate and can affect the outcomes of the experiment

  the role of variables is to measure or to manipulate
• what is the difference between a confounding and extraneous variable?

  An extraneous variable is any variable that you're not investigating that can potentially affect the dependent variable of your research study. A confounding variable is a type of extraneous variable that not only affects the dependent variable, but is also related to the independent variable.
• what is descriptive stats?

  • The notion of condensing and representing large data sets
  • Measures of central tendency and dispersion
  • Mean/std dev, median/IQR etc

  
• what is the difference between inferential and descriptive stats?

  descriptive statistics state facts and proven outcomes from a population, whereas inferential statistics analyze samplings to make predictions about larger populations.
• examples of descriptive stats

  • linear regression
  • hypothesis testing
• what is the difference between a sample and population?

  • If we have the entire population, we can draw conclusions from the data

  population parameter

  • If we only have some of our population i.e. a sample, we can only make inferences

  Sample statistic
• what is important to consider when looking at the sample pop

  ▸sample should represent the population we draw from. We need to be confident we are drawing the correct conclusions i.e. Inferential Statistics
• what is the null hypothesis?

  ▸Null hypothesis = default position. Most reasonable, nothing going on

  No difference, no relationship.

  “Innocent until proven guilty” - conservative
• what is the difference between a type I and type II error in hypothesis testing?


  • Type I Error (false positive)
  • Type II Error (false negative)
• what is the p-value?


  • the probability under the null hypothesis, of obtaining a result equal to or more extreme than what was actually observed.
  • between 0 and 1
  • 0 = very unlikely to be due to chance, due to bias, reason or relation
• when can the null hypothesis be rejected?


  when the p value is low
• how is the p-value normally set?

  ▸p-value low, it’s reasonable to reject H0

  ▹Evidence from our data supports the decision▹Never prove/disprove. – no binary “correct”

  ▹Usually set α to 5% or p < 0.05

  ▹report actual p values.. p = ?, p < .001

  • Arbitrary but agreed upon
  • Statistical Significance
• what is alpha in stats?

  the level of significance
• errors in hypothesis testing -> draw the table

  here

  
• a graphical representation of hypothesis testing - errors

  here

  
• what is beta in statisitics?

  The beta level (often simply called beta) is the probability of making a Type II error (accepting the null hypothesis when the null hypothesis is false). It is directly related to power, the probability of rejecting the null hypothesis when the null hypothesis is false.
• what is the relevance of the effect size?


  here

  
• ▸In a single sample, you either contain the population parameter or not we never know which it is ▸However…

    …in the long run, 95% of confidence intervals will contain the true population parameter.
• what is a confidence interval?

  A confidence interval is a range of values, bounded above and below the statistic's mean, that likely would contain an unknown population parameter.

  A confidence interval is a range around a measurement that conveys how precise the measurement is. 
• ▸Capture percentage = percentage of future sample estimates that fall within an observed CI

  ▹Roughly 83.4% of the time.

  ▹Depends on the CI that you happened to observe.

  ▹Contrasted with credible intervals in Bayesian stats
• what does it mean if Confidence intervals overlap?

  there is no real difference, not significant, wider interval = more uncertain
• list 5 types of research design

  • descriptive
  • correlational
  • experimental
  • review
  • meta-analytical
• what is descriptive research design?


  • case study
  • naturalist observation
  • Cannot statistically infer unless we have the entire population at which point you are no longer inferring…
• what is correlational research design?


  • case control, observational
  • considers the relationship between two variables free from manipulation
  • cannot determine cause and effect (third variable problem, spurious correlations)
• how can the relationship between two variables in correlational research design be analuysed?

  • spearman's
  • pearsons
  • point biserial
• what happens if there is a third factor in a correlational research design

  ▸Potentially more than two

  ▹Can be predictive and forecast

  ▹Modelling a “line” to your data

  ▹Linear regression, multiple regression, non-linear, multi-level model

  ▹Does this by explaining how much variance the model explains (r2)
• Research Design - experimental


  what are the key parts?

  
• ▸Review

  ▹Literature review

  ▹Overview of previously published works

  ▹Generally descriptive

  ▹Systematic review

  ▹Similar but with detailed and comprehensive plan/search strategy
• ▸Meta-analysis

  ▹Analysis that combines findings of multiple studies

  ▹Contrast results across studies

  ▹Provide better estimate of the unknown population effect size
• what is a Z test?

  A z-test is a statistical test used to determine whether two population means are different when the variances are known and the sample size is large.

  It can also be used to compare one mean to a hypothesized value.

  The data must approximately fit a normal distribution, otherwise the test doesn't work.