Segars: Introduction to Biostatisitics Flashcards
What is a Null hypothesis (H0)
-there will be NO change!!!!!
What is aternative hypthesis (H1)?
-there WILL be a change between groups
What does it mean when something is normally distributed?
-when mean, median, and mode are all equa
% found withing 1 SD of the mean on a normally distributed curve?
-68%
2 SD’s
-95%
3SD’s
-99.7%
What does a + skew curve look like?
- tail pointing to right
- Mean>median
- mode will always be that peak
negative skew
- tail pointing to left
- Mean< median
What is skewness?
-a measure of the asymmetry of a distribution
What are the 3 kinds of statistical test?
- Nominal
- ordinal
- Interval
Nominal
- no magntitude
- no consistency of scale
- named… only 2 things
Ordinal
- yes magnitude
- no consistency of scale
- ex: pain scale
- +3 categories
Interval/ratio
- yes magnitude
- yes consistency of scale
- ex: age
What are the required assumptions of interval data?
- normally distributed
- equal variances :use levene’s test
- Randomly-derived and independent
1 of the 4 key questions to selevting the correct statistical test?
- What DATA LEVEL is being recorded
- is it Nominal, ordinal, or interval???
2
- what type of comparison/assessment is desired?
- Correlation
- regression
- survival comparison
- group comparison
What are the values for correlation?
-+1 and -1, upwards slope and downwards slope (45 degrees)
If we are looking for a correlation test, what are the possibilities?
- Nominal= contingency coefficient
- Ordinal= Spearman correlation
- Interval= pearson correlation
What do we have to know about pearson correlation?
-just assesses for linear correlation, there may still be non linear correlations present if pearson correlation non-signif.
After we decide if it is nominal, ordinal, or interval, what do we do next?
- look for 5 things
- 2 groups
- > or = 3 groups
- Proportion of events (survival)
- Measure of correlation
- Prediction or association
What are survival tests commonly represented by?
-a kaplan meier curve
What will all survival tests have on the bottom?
-time on x axis
If it is survival, what are the possibiliities?
- Nominal= log rank test
- Ordinal= cox-proportional hazards test
- interval= Kaplan Meier test
what is a regresssion?
- provides a measure of the relationship between variables by allowing the prediction about the dependent, or outcome, variable knowing the value/category of independent variables (IV’s)
- also able to calculate OR for a measure of association and control for confounding IVs
If it is outcome prediction/association (or regression), what are the possibilites?
- Nominal= logisitic regression
- ordinal= multinomial logistic regression
- Interval= linear regression
What are the last 2 questions we have to ask before we pick which statistical test we want to do?
- How many groups
- is the data independent or related
Nominal 2 groupd of indy data
-pearson’s chi square test
Nominal >=3 groups indy
-chi square test of independence
> = 2 groups with expected cell count of >5?
- fisher’s exact test
- aka, not a lot of people
What are the key words for “paired” or “related” data?
- pre vs post
- before and after
- baseline vs end
Ordinal and 2 groups of indy data?
-Mann-whitney test
> = 3 groups of indy data
-Kruskal-Wallis test
What does the Mann whitney and kruskal wallis tests compare?
-the median values between the groups
2 groups of pair/related data
-wilcoxon signed rank test
> =3 groups of paired/related data
- Friedman test
- both compar median values again
post hoc tests for 3 or more gorup comparisons
- student newman keul test
- dunnett test
- dunn test: good for when groups are NOT of equal size, we need that for the other 2….
Interval 2 grops of indy data
-student t test
Interval >=3 groups of indy data
- Analysis of variance (ANOVA)
- both tests compare the MEANS
- Multiple analysis of variance (MANOVA)
Interval >= 3 groups of independent data w/ confounders?
- analysis of Co variance (ANCOVA)
- MANCOVA
Interval, 2 groups, paired/related
-Paired t-test
> =3 groups, interval, related
- repeated measures ANOVA
- Repeated measured MANOVA
Interval, >= 3 groups, paired, confounders
- repeated measures ANCOVA
- Repeated measures MANVOVA
Interval data post hoc tests for 3 or more group comparisons
- student-newman-keul test
- Dunnett test
- Tukey or Scheffe tests
- Bonferroni corretion
- Dunn test: rememebr, NOT EQUAL SIZE FOR THIS ONE, BUT YES EQUAL FOR THE OTHER 2
What are the 4 questions we have to ask when selecting the correct statistical test?
- What data level is being recorded?
- What type of comparison is desired?
- how many groups?
- Indy or related?
What is a type 1 error?
- rejecting a null hypothesis when it is actually true and YOU SHOULD HAVE ACCEPTED IT YOU DUMB ASS!
- insert picture guys that look the same but we said that they are different
- same, yet we said they were different
What is a Type 2 error?
- accepting a null hypothesis when you should have rejected it
- insert pic of a fat guy next to a skinny guy and we said that they were the same… different, yet we said it was the same
What is power (1-B error)?
-the ability of a study design to detect a true difference if one truly exists between group comparisons, and therefore the level of accuracy in correctly accepting/rejecting the null hypothesis
What is the relationship between sample size and power?
-the larger the sample size, the greater the likelihood of detecting a difference if one truly exists
What 3 things go into sample size determination?
- minimum difference between groups deemed relevant/important
- expected variation of measurement
- alpha and beta error rates and confidence interval
What is the p value
-probability of making a type 1 error
What is the magic number for p value?
- 5% or 0.05
- if below that, then we can go ahead and reject the null hypothesis
What is the other thing that we use besides p value that is based on variation in sample and sample size?
- Confidence Interval (CI)
- 95% was blue
When are the stats NOT significant for the CI?
-if CI crosses 1.0 (for ratios) or 0.0
