We’re skipping section 4.1 on percentile ranks – conceptually, it’s very similar to the formula used for calculating the median when there is replication of the middle value.
Standard Scores (4.2)
- Standard Scores provide a method for interpreting single scores within the set or distribution
- Chapters 2 and 3 focused on statistical indices for describing an entire set of scores
- Standard Scores can tell us how much different one person’s score is from another’s.
- A standard score is computed by taking into account how far a given score deviates from the mean.
- The deviation from the mean is expressed in terms of standard deviation units
- This is done by computing: standard score =
- This is the deviation score (for that specific case) divided by the standard deviation for the distribution
- Standard score tells you the number of standard deviation units above or below the mean a particular score is
EXAMPLE: Back to the set of scores for how old you think I am.
Score (X) – Raw Scores
23 We previously computed
24
24 M = 25.67
24 SD = 2.32
25
25
25
25
26
27
28
32
- What is the standard score for the raw score of 27?
Standard score = (.57 SD’s above the mean)
- What is the standard score for the raw score of 23?
Stand. score = (1.15 SD’s below the mean)
- What is the standard score for the raw score of 28?
Stand. Score = (1 SD above the mean)
Properties of Standard Scores
- Sign
- Negative – score is below the mean
- Zero – score is equal to the mean
- Positive – score is above the mean
- The mean of any set of standard scores will always be zero ()
- The variance and SD of any set of standard scores will always be 1.00
**If you turned all your raw scores into standard scores, then calculated , , and , then , and
Uses of Standard Scores
- Standard scores allow us to make comparisons between scores from distributions that have different means and standard deviations
** Remember, when we convert to standard scores, M = 0 and SD = 1.0, so it is possible to compare two scores that were originally from two different distributions.
EXAMPLE: You give two tests.
Verbal Ability Quantitative Ability
100 questions 100 questions
X = 85 X= 95
M = 65 M = 80
SD = 8.00 SD = 13.00
std. score = std. score =
- Looking at the raw scores, it looks like this person did better on the Quantitative Ability test.
- But when take into account how others did on these tests, this person actually did better on the verbal ability test
Standard Scores and the Normal Distribution (4.3)
- Normal Distribution (Chapter 2: 2.9)
- Normal distributions have the following properties:
- Are bell shaped
- Are symmetrical around the mean
- Have Mean = Median = Mode
- There is a different normal distribution for every unique combination of M and SD, but all have the 3 properties above
- When convert all scores to standard scores, there are some known properties of standard scores and the normal distribution.
- In any normal distribution, the proportion of scores falling above or below a given standard score is always the same.
- Example: 50% of the scores always occur above or below a standard score of zero (i.e., the raw score equals the mean) for any normal distribution
- The proportion of scores that occur between two specific standard scores is the same for all normal distributions
- Example: The same proportion of scores will occur between a standard score of 1 and a standard score of 2 for all normal distributions.
- Discussion of Figure 4.1 – p. 110
- .3413, or about 34% of the scores fall between standard scores of 0 and 1(or 0 and -1)
- .6826, or about 68% of scores fall between standard scores of –1 and +1
- .9544, or about 95% of scores fall between –2 and +2
- .9974, or about 99% of scores fall between –3 and +3
- When we are talking about standard scores in a normal distribution, we call the standard scores z-scores
- Is computed the same way as before:
- Appendix B (p. 562-571) gives proportions falling under more specific proportions of the normal curve than are presented in Figure 4.1 (for many more z scores than the whole numbers listed in Figure 4.1)
- This table gives us the following information:
- Column 1 – absolute value of z scores, starting w/.00 through 4.00
- Column 2 – the proportion of scores falling between any positive z score and its negative value
- Example: Recall from Fig. 4.1, the proportion of scores between z’s of –1 and +1 is .6826 or 68.26%
- Example: look at z=1.96 – the proportion of scores between z’s of –1.96 and +1.96 is .9500 or 95%
- Column 3 – the proportion of scores falling above a positive z or below a negative z
- Example: look at z = 1.96 --- .0250 or 2.5% of scores fall above +1.96 and 2.5% fall below –1.96
- Column 4 – the proportion of scores above its positive value and below its negative value (in other words, Column 3 multiplied by 2)
- Example: look again at z = 1.96 --- .0500 or 5% of scores fall above +1.96 and below –1.96 combined
- Column 5 – the proportion of scores between any z score and the mean
- Example: look again at z = 1.96 --- .4750 or 47.5% of the scores fall between zero and +1.96 (or between 0 and –1.96)
- Notice that column 2 shows these two proportions combined -- .9500
- Many Normal Distribution tables only give you three columns (1, 3 & 5), but because of the symmetry of the normal distribution, these are the only columns that you NEED.
- Table is used to find proportions of scores in a given range under the normal curve. This is used when we have converted our raw scores to z scores (standard scores)
- If we can assume a set of scores is normally distributed, we can use the table to find the percentage of scores in a given portion of the normal distribution.
- The z score of +/- 1.96 turns out to be very important in psychological research
- We will see this importance when we get into inferential statistics
- Know that 95% of scores fall between z’s of –1.96 and +1.96
- How do we use Appendix B for an actual problem?
- For example, assume that verbal GRE scores approximate a normal curve with a mean of 430 and a standard deviation of 100.
- What proportion of scores are greater than 500?
a. Convert to z score: (500 – 430)/100 = .70
b. Go to Appendix B – Column 3
c. .2420 (or 24.2%) of the scores are greater than 500
- What proportion of scores are lower than 445?
a. Convert to z score: (445 – 430)/100 = .15
b. Go to Appendix B – No one column will answer this!
c. Column 5: .0596 (proportion between .15 and mean)
d. We know that .5000 of scores below the mean
e. .0509 + .5000 = .5509
f. .5509 (or 55.09%) of scores are lower than 445
- What proportion of scores are between 450 and 470?
a. Convert 450 to z score: (450-430)/100 = .20
b. Convert 470 to z score: (470-430)/100 = .40
c. Go to Appendix B – Great, Now what?
d. Column 3 for .20: .4207; Column 3 for .40: .3446
e. Subtract smaller proportion from larger proportion: .0761
f. .0761 (or 7.61%) of scores between 450 and 470
- What proportion of scores are between 400 and 450?
a. Convert 400 to z score: (400-430)/100 = -.30
b. Convert 450 to z score: .20
c. Go to Appendix B – But this isn’t the same as above!
d. Column 5 for -.30: .1179 (remember, Column A is the absolute value)
e. Column 5 for .20: .0793
f. Add proportions together (because on either side of mean)
g. .1972 (or 19.72%) of scores between 400 and 450
- What proportion of scores are less than 400 and more than 500?
a. Convert 400 to z score: -.30
b. Convert 500 to z score: .70
c. Go to Appendix B – This is getting easier, but…
d. Column 3 for -.30: .3821; Column 3 for .70: .2420
e. Add proportions together
f. .6241 (or 62.41%) of scores are less than 400 and more than 500
- Now you see why many tables only have columns 1, 3 & 5 – rarely do you actually use columns 2 & 4 and you can easily get that information by doubling the values of column 3 (equals column 4) and doubling the values of column 5 (equals column 2)
- Remember: It is often best to sketch a normal distribution and then shade in the section you are trying to find. This will help you determine the computations you need to do in order to answer the question.
- Remember: To convert proportions into percentages, simply multiply by 100. Often easier to talk in terms of percentages than proportions.
Standard Scores and the Shape of the Distribution (4.4)
- The process of standardizing scores does not change the fundamental shape of the distribution
- a positively skewed set of scores will remain positively skewed when standardized
- a platykurtic set of scores will remain platykurtic
- Standardization affects the central tendency and variability of the scores but not the skewness or kurtosis of scores