Statistics

Question 1

Positive skew

Answer 1

skewed to the left

Question 2

negative skew

Answer 2

skewed to the right

Question 3

what is kurtosis

Answer 3

Pointy/flat. How heavily the tails of a distribution differ from the tails of a normal distribution

Question 4

what is a hypothesisa

Answer 4

a predictable theory to test

Question 5

what is a alpha level

Answer 5

the significance level

level which you would be happy to make an error rejecting the null hypothesis if it were really true

Question 6

what does it mean if If p < alpha

Answer 6

null hypothesis is rejected this is a statistically significant result

Question 7

what is a type 1 error

Answer 7

• False positive
• There wasn’t really a difference/ relationship but you said there was
• You rejected the null hypothesis when it was actually true
• A higher alpha level = more type 1 errors (the acceptable level of error is higher and therefore you will make more errors)
• Also depends on the type of stats you use and sample
Alpha too high

Question 8

what is a type 2 error

Answer 8

• False negative
• There was a difference/ relationship but you said there wasn’t
• You accepted the null hypothesis and missed the difference/ relationship
• A lower alpha level = more type 2 errors (acceptable level of error is low and therefore you will miss differences/ relationships because you think these are errors)
• Also depends on the type of stats you use and sample
Alpha too low

Question 9

what is power

Answer 9

the chance of spotting an error

A powerful test is one that spots the difference if there is one

Question 10

what are Parametric tests

Answer 10

• Interval/ ratio scale, data normally distributed, data from multiple groups have the same variance, data are independent
• More powerful
• Greater range of tests
Easier to understand and compute

Question 11

give the equation for the regression line

Answer 11

Y=mx + c where m is gradient, c is intercept on the y axis

Question 12

what is r

Answer 12

R = 0 means there is no correlation

Question 13

what is r squared

Answer 13

R squared is a good indicator of goodness of fit

Question 14

what is regression

Answer 14

• Regression puts the line through and gives the equation so you can calculate y
• Regression is basically the same thing as correlation but with more information
• Linear regression describes strength and direction of the relationship between 2 variables where you measure one and manipulate or control the other variable as basis for prediction
• —-> Bivariate regression
• Linear regression helps to establish the correlational relationship
• Residual = the difference between a predicted score and an actual score. Residuals are - if the line is perfect (measure of error)
Bivariate regression - need to make sure residuals are normally distributed (Shapiro Wilk), homoscedasicity

Question 15

what is linear regression

Answer 15

• Describes strength and direction of the relationship between variables where you measure one and manipulate or control another
• Helps to establish the correlational relationship as a basis for prediction
• To check linear regression is significant, use the ANOVA table
When you run linear regression analysis in SPSS, Pearson’s correlation coefficient is calculated as part of the analysis

Question 16

what is multiple linear regressoin

Answer 16

• Normality of residuals 
	• At least 2 IV 
	• Each IV has at least 10 participants 
	• Linearity 
Normal distribution of all unstandardized residuals combined together

Question 17

what is linear bivariate regression

Answer 17

• Residuals normally distributed
• Only looking for linear relationship
• Need to make sure the residuals are not related to each other
One outlier can cause a big proble

Question 18

what is multiple regression

Answer 18

• When there are more variables

How several things are relate

Question 19

what is Standard error of the estimate

Answer 19

• Tells us how accurate your predicted value is
  
  • Used to build a 95% confidence interval
  • To calculate how much error there would be if you used the linear regression to predict a score
  • 95% confidence interval is +/- 1.96 x standard error of the estimate
  • Confidence interval= the range of the scores what the y will be in because the line doesn’t fit exactly and therefore we might be slightly off with our prediction
  • Confidence interval means that we can be 95% sure that the predicted scores will ie within the identified range of scores
  • To calculate the confidence interval, multiply the standard error by 1.96

Question 20

what are beta weights

Answer 20

• Allow you to compare the importance of different IV which are usually calculated using different measurement units
  
  • Standardised scores of your raw values which allow for easier comparisons to be made
  • The highest beta weight is the most important variable in your multiple regression
  • Absolute value = ignorewhether it is positive or negative

Question 21

what are error bars

Answer 21

Give you an idea how precise the results are (indicate error)

Question 22

what is the absolute difference

Answer 22

• Simply find a difference between the two means and ignore if it is positive or negative
• Percentage of overall mean usually asked in the context of absolute difference between the means of the 2 groups. Need to find the absolute difference between the means of 2 groups. You then need to find the overall mean of 2 groups and express as a percentage
Absolute difference % = absolute difference / overall mean x 100

Question 23

when would you use a one way between subjects ANOVA

Answer 23

• Interval/ ratio
  
  • At least 3 levels of IV
  • Normality of DV within every level of IV
  • Homogeneity of variance