Statistical Interference Week 6 B&R

0.0(0)
Studied by 0 people
call kaiCall Kai
learnLearn
examPractice Test
spaced repetitionSpaced Repetition
heart puzzleMatch
flashcardsFlashcards
GameKnowt Play
Card Sorting

1/15

encourage image

There's no tags or description

Looks like no tags are added yet.

16 Terms

1
New cards

Why do statistics use samples to make inferences of population?

  • Estimate is a known quantity calculated from the sample data 

  • Use estimate to try and calculate the unknown population- Parameter 

Sample is very unlikely to represent population perfectly- attach confidence and probability to our statements  

  • We use small groups to learn something the larger population represents 


2
New cards

What are the two types of statistical tests?

T-test: testing for difference between two groups/conditions 

Correlation: exploring relationship between two variables 

3
New cards

What are the main principles behind a t-test?

Is variation between-group variation greater than within-group variation? 

  • t-value= mean difference (signal)/ standard error (noise)  

High t value indicates a clear signal  

4
New cards

What are the two main types of t-test?

Independent- compares means of two independent groups  

Paired (dependent)- compare the means of two related group samples 

5
New cards

When do we use a paired t-test?

Before-and-after measurements, repeated measurements, two conditions applied to the same subject, matched individuals and two methods applied to the same item 

6
New cards

What is statistical interference?

Where statistics use samples to make ‘inferences’ about a population

  • ESTIMATE is a known quantity calculated from a sample of data

Use the estimate to try and determine the unknown population PARAMETER e.g. population mean

• Sample is very unlikely to represent the population perfectly – we attach probability or confidence to our statement

  • study small group to learn something larger population represents


7
New cards

What is Descriptive statistics?

Summarize and describes the data you have e.g. pie chart, bar graph

8
New cards

What is inferential statistics?

Makes conclusions about larger population e.g. hypothesis testing

9
New cards

What are the three levels of significance?

Statistical significance: Probability of obtaining the sample result if the null hypothesis was true. 

Theoretical significance: Finding is relevant for a theory or leads to a new experiment. 

  • Many results that are statistically significant don’t meet this criteria 

Clinical (practical) significance: A finding is relevant for diagnosing or treating a disease 

10
New cards

Why might group means be different?

  • There is a 'real' difference in population and the difference in the sample group reflects this- SIGNAL 

  • No ‘real’ difference the population and difference in sample reflects sampling in variation and error- NOISE 


11
New cards

What are type 1 and 2 hypothesis testing errors?

In hypothesis testing, a Type I error is a false positive

  • concluding an effect exists when it does not.

Type II error is a false negative.

  • failing to detect an effect that exists


12
New cards

How do we reduce type 1 errors?

Reduce α threshold and control multiple testing. More likely that effect is real (decrease Type I errors) but harder to find real effect. 

13
New cards

How do we reduce type 2 errors?

Increase power, often through a larger sample or better study design. Usual to set power around 80%. 

14
New cards

What are the features of correlation?

  • Correlations only examine linear relationships 

  • Correlation is independent of units  

  • Used if both variables are measured (not manipulated) 

Significance is highly dependent on sample size  

  • Very small correlation coefficients can be statistically significant if there is a very large sample 

  • Consider practical importance of findings   


<ul><li><p class="Paragraph SCXW266062611 BCX0" style="text-align: left;"><span style="line-height: 20.925px;">Correlations only examine linear relationships&nbsp;</span></p></li></ul><ul><li><p class="Paragraph SCXW266062611 BCX0" style="text-align: left;"><span style="line-height: 20.925px;">Correlation is independent of units&nbsp;&nbsp;</span></p></li></ul><ul><li><p class="Paragraph SCXW266062611 BCX0" style="text-align: left;"><span style="line-height: 20.925px;">Used if both variables are measured (not manipulated)&nbsp;</span></p></li></ul><p class="Paragraph SCXW266062611 BCX0" style="text-align: left;"><span style="line-height: 20.925px;">Significance is highly dependent on sample size&nbsp;&nbsp;</span></p><ul><li><p class="Paragraph SCXW83487785 BCX0" style="text-align: left;"><span style="line-height: 20.925px;">Very small correlation coefficients can be statistically significant if there is a very large sample&nbsp;</span></p></li></ul><ul><li><p class="Paragraph SCXW83487785 BCX0" style="text-align: left;"><span style="line-height: 20.925px;">Consider practical importance of findings&nbsp;&nbsp;&nbsp;</span></p></li></ul><p></p>
15
New cards

How do we choose statistical data if normally distributed?

Is the data normally distributed? 

Yes- parametric tests (more powerful) 

No- nonparametric tests (ranking of data) 

16
New cards

What are non-parametric tests?

Nonparametric tests- used with data that isn't normally distributed  

  • T statistics are calculated using rankings  

  • Less powerful but more robust 

Example- spear man's rank correlation