Justin's + 2018 Stats Flashcards
When would you use a one tailed v two tailed test?
When the alternative hypothesis only goes in one direction rather than two
What variable is not used for power calculation?
a. actual means for groups
b. expected difference
c. significance level
a. actual means for groups
Power = 1-beta; beta is the probability of a type 2 error
*don’t totally understand this question
When conducting a power calculation, you have to take into account:
- how much of a difference you expect to see (ie a very big difference or a little difference, as this affects the required sample size)
- what level of significance do you want (setting your alpha and beta)
For this question, you wouldn’t know the means yet. Power calculation is your set up to an experiment. Comparing means is in an ANOVA test.
What is the measure of intra-observer variability
Kappa
What factors are important for power calculation?
Effect size and sample size
What does the receiver-operator curve (ROC) measure?
x axis= false positive rate (1-specificity)
y axis= true positive rate (sensitivity)
The true positive rate (sensitivity) is plotted in function of the false positive rate for different cut-off points. Each point represents a sensitivity/specificity pair corresponding to a particular decision threshold.
Want to be as close to upper left of curve as possible, higher overall accuracy.
What is type 2 error?
Type II error: failing to reject a false null hypothesis. (“false negative”)
Beta
Power=1-beta
type I error is the mistaken rejection of an actually true null hypothesis (“false positive”)
What is type 1 error?
Erroneously rejecting the null hypothesis
Alpha
What curve and analysis for survival analysis?
Kaplan-Meier
Log rank (only if simple variable–compares two drugs and produces a p-value, however does not provide the magnituted of the effect)
Cox proportional hazards (can quantify the effect of multiple variables on survival).
What are statistical tests used for evaluation of independent variables?
ANOVA (for dependent or independent)
Mann-Whitney U
Unpaired T test
What is a statistical test to evaluate continuous variables in normally distributed population?
Unpaired t-test
What is the statistical test that allows you to compare three groups?
ANOVA
What are statistical tests for comparing nominal (categorical) variables in a normally distributed population?
Chi squared
What is the statistical test to compare before and after intervention?
PAIRED t-test
Positive predictive value (PPV) is affected by what?
Prevalence
Higher prevalence will increase PPV and decrease NPV–no impact on sensitivity or specificity
Likelihood ratios do not depend on prevelence
Match the following parametric tests with their non-parametric counterpart:
Parametric
Paired t-test
Unpaired t-test
Pearson correlation
One way ANOVA
Non-parametric
Mann-Whitney U test
Kruskal Wallis test
Wilcoxon Rank sum test
Spearman correlation
Parametric–>Non-parametric
Paired t-test–>Wilcoxon Rank sum test
Unpaired t-test–>Mann-Whitney U test
Pearson correlation–>Spearman correlation
One way ANOVA–>Kruskal Wallis test
You conduct a clinical trial and get p<0.01. The following are true except?
a. study was significant
b. smaller sample size may have resulted in non-significant finding
c. reject the null
d. you didn’t have enough power
d. you didn’t have enough power
Aside from randomization, how can you control for confounding variables?
Multivariate logistic regression
How can you check the impact of an independent variable?
Logistic regression
What is the formula for odds ratio (OR) v relative risk (RR)?
When do you use each?
OR=(a/b)/(c/d) or (axd)/(bxc)
RR=[a/(a+b)] / [c/(c+d)]
OR for case-control–compares presence/absence of exposure knowing the outcome
RR for cohort study–know exposure status, then calculate probability of an event
What test allows you to check the effect of multiple variables on survival?
Cox proportional hazard
Calculation of positive predictive value, negative predictive value, sensitivity and specificity–what are their definitions?
Sensitivity=true positive
Specificity=true negative
PPV=true positive/test positive
NPV=true negative/test negative
What is the best test for a case-control study?
Odds ratio
“Case is odd”
What is the best randomization method?
a. simple (coin flip)
b. block
c. alternate assignments
b. block
This method achieves balance in sample size
Alternate assignments should not be used
What is the best initial test to start with to evaluate smoking and certain type of cancer?
a. prospective randomized control trial
b. case control
c. cohort study
d. chart review
b. case control
What is the best initial test to assess the relationship between prenatal vitamin and ovarian cancer?
a. case control
b. cohort
c. meta analysis
d. randomized control trial
a. case control
What test allows you to compare the mean among three groups?
ANOVA
What are the axes on a receiver-operator curve (ROC)?
x=1-specificity
y=sensitivity
What is positive predictive value dependent on?
Prevalence
What is the difference in number of people who get a disease exposed to a risk minus the people with a disease not exposed to a risk?
a. attributable risk
b. absolute risk
c. risk difference
a. attributable risk
What affects the sample size in clinical studies?
Factors affecting sample size are: study design, method of sample and outcome measures–effect size, standard deviation, study power, and significance level
What is quality assurance?
These activities provide confidence a service will fulfill quality requirements
What is quality control?
This is the part of quality management that focuses on fulfilling quality requirements
What is a basket trial?
Trial that tests 1 drug against 1 mutation in many cancer types, can increase participant numbers
*A type of clinical trial that tests how well a new drug or other substance works in patients who have different types of cancer that all have the same mutation or biomarker. In basket trials, patients all receive the same treatment that targets the specific mutation or biomarker found in their cancer.
What is an umbrella trial?
Many arms within one trial (evaluate multiple moleculartly guide therapies), participants assigned based on tumor mutation specifics or mollecular profiles of tumors.
What is the standard error in relation to the standard deviation?
The standard error equals the standard deviation divided by the square root of the sample size.
It gets smaller as the sample size increases.
What is the definition of standard error?
The standard deviation divided by the square root of the sample size
Tells you how different the population mean is likely to be from the sample mean
Standard deviation measures the variability from specific data points to the mean. (One experiment)
Standard error of the mean measures the precision of the sample mean (our one experiment) to the population mean that it is meant to estimate (repeat the experiment with different groups ten times and take the mean) aka the mean of the mean
what is the best test for case control?
Odds ratio = (a/b) / (c/d) = ad/bc
*The odds of the event occurring in an exposed group versus the odds of the event occurring in a non-exposed group. It helps identify how likely an exposure is to lead to a specific event.
When do you use logistic regression?
Control for confounding variables
*Use logistic regression when you expect a binary outcome (for example, yes or no).
What is the test for nominal variable that is normally distributed?
Chi square
What is the test for means for two groups of people before and after an intervention?
Paired t-test
What is the receiver-operator curve (ROC) best for?
Accuracy
Y axis sensitivity
X axis 1 - specificity
What type of error is it when you inappropriately reject the null hypothesis?
Type 1
How does the choice of statistical test affect power?
Non-parametric tests have less power for the same sample size compared to the corresponding parametric test.
*When SHOULD you stick with a nonparametric test:
- Your area of study is better represented by the median.
- You have a very small sample size and non-normal looking data.
- You have ordinal data, ranked data, or outliers that you can’t remove.
What is the test to compare multiple means?
ANOVA (analysis of variance)
What is the best initial study design to look at smoking and cancer?
Case control
What is the effect of increasing selection size in an observational study?
Decreases selection bias
*I did not find anywhere that larger sample sizes decrease selection bias. The larger the study sample size, the smaller the margin of error. Larger sample sizes allow researchers to control the risk of reporting false-negative or false-positive findings. The greater number of samples, the greater the precision of results will be.
What has the biggest effect on positive predictive value?
Prevalence
What is the best test to compare the mean for normally distributed data?
t- test
*if less then two groups being compared
