Chapter 6: Correlation and Predicted Variance Flashcards Preview

Statistics > Chapter 6: Correlation and Predicted Variance > Flashcards

Flashcards in Chapter 6: Correlation and Predicted Variance Deck (30)
Loading flashcards...
1
Q

Relationships in statistics are looked at for what three reasons?

A
  • comparison of different distributions - determining causality - psychometric properties of questionnaires
2
Q

Correlations? (4)

A
  • interested in relationship not magnitude of one over the other.. - one variable carries information about another variable - easy first step in determining causality but there has to be a correlation - Correlations do not = CAUSATION. ever.
3
Q

When constructing a scatter what is being graphed?

A

X and Y data points….

4
Q

Linear relationships? What the heck are they?

A
  • straight line…what we are interested in!
5
Q

Curve relationship?

A
  • not interested in this - BAD…..they are bias..
6
Q

Describe a positive relationship?

A
  • each variable is increasing together - correlation is evident - direct
7
Q

Describe a negative relationship?

A
  • one variable is increasing while the other is decreasing… -inverse
8
Q

Describe a perfect relationship ? Are they common?

A
  • all data points fall exactly on the line - not very common
9
Q

Describe an imperfect relationship? Are they common?

A
  • all data points do not fall on the line - still linear -very common -line of best fit/ regression line L> basically a mean….points around it = variance…
10
Q

Correlation part 2 yo

A
  • family of statistical tests that quantify the relationship between the variables….
11
Q

Correlation coefficient what the heck is that?

A

single # that summarizes the relationship of two variables.. ranges from +1.00 to - 1.00 L> signs only indicate direction….both are equally as strong correlations

12
Q

Characteristics of Correlation coefficient? most common correlation coefficient values?

A
  • zero = very weak/ no relationship - + correlation = with every one unit increase there is a proportional increase in another variable..and vise versa - -0.5- + 0.5
13
Q

Describe the Pearson r Correlation Coefficient.. how does it get around the issue of varying units?

A
  • extent that paired scores occupy the same or opposite positions within their distributions…. - convert data into z score…so no unit score issue…
14
Q

What is the raw score Pearson r Formula?? What does each part represent?

A
15
Q

Variability of Y can be explained by?

A
  • X
16
Q

When r= 0 the best predictor of Y is the _____ of the y scores.. What are the erros associated with this?

A
  • mean, imperfection in prediction…
17
Q

When r does not equal 0 the best predictor of Y is ____. Prediction errors?

A
  • X
  • when x is a predictor the error goes down significantly
18
Q

The total deviation of score is divisible into two parts..what are they?

A
  1. the distance from the regression line to the mean line = deviation accounted for by X (A)
  2. The distance from the regression line to the point in question….prediction error (B)

A+B= deviation of score

19
Q

The deviation of Yi

is what??

A

prediction error + deviation of Yi accounted for by X

20
Q

The total variablilty of Y ?

A

Variability of prediction errors + variability of Y accounted for by x

21
Q

When correlation goes up the variability of prediction errors goes _____, variability of Y accounted for by Y _____.

A

-decreases, increases

22
Q

When r= 0 the variability of P.errors = ________. Variability of Y accounted for x =___?

A

-min variable, none

23
Q

When r= 1.00 Variability of P.errors = _____ and the variability of Y accounted for by x ____.

The greater x is = ___ proportion of Y is accounted for

A
  • none, maximum
  • greater
24
Q

Explained variance? Explain it ! (4)

A
  • r= correlation coefficient

L> magnitude and direction of relationship

  • r = coefficient of determination

L>proportion of total variability in Y accounted for by x…

25
Q

Describe the Explained Variance table…

A

r r2

0.10 1

.20 4

  1. 30 9
  2. 40 16
  3. 50 25
  4. 60 36
  5. 70 49
  6. 80 64
  7. 90 81
  8. 00 100

1, 9 and 16 are the most common in the behavioural sciences

25, 36 are large correlations

49, approx 1/2 variance

64, 81 and 100 are rarely seen ever…more so in psychometrics..

26
Q

Issues with a curve shaped graph?? GO

A

r values cancel each other out

magnitude of correlation is severly reduced….

ex: performance in anxiety

27
Q

Describe the following coefficients !

Pearson r

Biserial rb

Spearman rank order rho

Phi ro

A
  • 2 interval/ or ratio
  • one interval/ratio and one dichotomous
  • 1 or both = ordinal
  • 2 dichotomous
28
Q

Phi coefficient??

|A|B|

|C|D|

formula?

A

AD-BC/ >/ (A+B)(C+D)(A+C)(B+D)

>/ = square root

29
Q

With the phi coefficient everything is related to ____

A

A

30
Q

What are the three issues with correlations?

A
  • restricted ranges : reduces magnitude of correlation…reducing variability
  • outliers: increases magnitude
  • correlation does not equal causation