Greife auf kostenlose Karteikarten, Zusammenfassungen, Übungsaufgaben und Altklausuren für deinen Statistics Kurs an der University of Pretoria zu.

- 4036 Karteikarten
- 955 Studierende
- 7 Lernmaterialien

Q:

Types of variables

A:

1.Continuous

*there's meaning between the two spectrum of the variables.e.g. age/weight/height -2.5 years means something

2.Categorical

*there's no meaning between the two.its either or

E.g.age GROUP,gender,citizenship

Also called discrete. If there's only 2 categories its dichotomous variables(e.g. gender)

Continuous variables can be changed to categorical e.g. age to age groups

Changing a non dichotomous variable to a dichotomous one is called "to dichotomize"

Changing a non dichotomous variable to a dichotomous one is called "to dichotomize"

Q:

What is the P value

A:

P value calculates the number under the curve where you plot the frequency of a certain result(frequency at which certain results occur) when you run a tests multiple times and plot it.it is the number under the curve that tells you whether (or not) its probable that the null hypothesis is true.

The probability that you rejecting the null hypothesis was done incorrectly.

If p value is less than 0.05(alpha value=5%) you can confidently REJECT the null hypothesis.that means you can confidently say there is a relationship between two variables!"statistical significance"

E.g. p value of 0.02 means theres a 2% chance that you incorrectly rejected the null hypothesis.2% chance that the null hypothesis is right.

If the p value is low,the null hypothesis must go!

So you want a low p value.low is p value less than alpha value.low p value means you have found something(your results) that has a statistical significance and is not due to chance

Affected by sample size.Larger samples give more correct P values.

Range = 0 - 1(0-100%)

Small P value suggests narrow CI (good)

Higher P value suggests wider CI

Q:

What is relative risk

A:

Its a comparison - the risk of outcome when you are exposed(absolute risk) vs the risk of outcome when you are not exposed

Q:

Types of bias

A:

1.Selection Bias

1.1.Healthy worker bias

1.2.Berkeson bias

1.3.Lost to follow up bias

1.4.non response bias

2.Information Bias - the means of taking info from subjects inadequate

2.1.Misclassification bias (when the records are wrong,assigning disease where its not there).can be differential and non differential

2.2.Recall bias (participants dont remember everything correctly)

2.3.Interviewer bias (Interviewer behavior affects respondent response e.g.tone of question)

3.Response bias

Respondee wants to give you answer they think you want ,not because of anything interviewer does

4.Detection bias

*surveillance

When you find something because you went looking for it.you are more likely to find it where you are looking for it.not that it wasnt there before e.g.seeing a high prevalence for a disease because we now have increased testing

5.Hawthorne/Rosenthal/Golem effect

**biases that results from the interaction of researchers with subjects.prevented by blinding in RCT studies

Hawthorne : people are likely to do better when watched.so more compliant patients are more likely to be chosen in trials(they adhere better because they are in a trial being watched)

Rosenthal : positive self fulfilling prophecy

Golem effect : negative self fulfilling prophecy

***Confounding

Can show a relationship where it doesn't exist or mask one that exists. Confounder creates illusion.

Confounders : age,gender,smoking.

Technically not bias.

Effect modification

Changes the direction/nature of a relationship that's real

Q:

What is mean

A:

Addition of all the variables divided by total number of variable.

Works only in normally distributed data.

Q:

What is a confidence Interval

A:

Another way to express that your results have a statistical significance.(like p value)

Range in which your value/results has a statistical meaning/significance.Preferred to a p-value,only use one of the other.

Point estimate - what we are caculating/actual parameter

E.g.average age of university students is "21years".21 years is the point estimate.then range of 18 to 21.5 years is where 95% of the times you will find the accurate average age range of varsity students(its your confidence interval)

Usually 95% CI.same as alpha value(its set)

CI gives us the range of values in which the actual answer probably sits.

Narrow CI - precise estimate or risk ratio/odds ratio.its good

Wide CI - imprecise estimate of rr/or.its bad.

If CI doesnt include 1(when using RR) its statistically significant

If CI doesnt include 0(when using RD) its statistically significant

Q:

Specificity calculates true negatives

A:

Specificity = TN÷(TN+FP)*** TN+FP gives total of negatives

Q:

What is attributable risk

A:

How much of the outcome was due to the exposure/risk factor

Q:

Define numbers needed to treat

A:

The number of patients who need to receive the intervention/exposure to prevent 1 outcome from happening

E.g.if outcome is death,the number of patients who need to be treated to prevent 1 death

Formula = 1/risk reduction(write in positive value even if result is negative)

Q:

Correction for selection bias

A:

Randomization

Restrictions

Adjustments

Matching

Multiplication

Stratification

Q:

What is median

A:

The middle number of all the data if it is organized in a numerical order.

Better used in non normally distributed data

Q:

PICO method of asking research questions

A:

Population group

Intervention

Control group

Outcome

Greife kostenlos auf tausende geteilte Karteikarten, Zusammenfassungen, Altklausuren und mehr zu.

Jetzt loslegen