PROBABILITY

KULIAH 8 OLEH ISMAIL KAILANI

1

INTRODUCTION TO PROBABILITY
The idea of research studies begin with a general question about the entire population, but the actual research is conducted using a sample. In this situation the role of inferential statistics is to use the sample data as the basis for answering questions about population. To accomplish this goal, inferential procedures are typically built around the concept of probability. Specifically, the relationship between samples and populations are usually defined in term of probability.
2

INTRODUCTION TO PROBABILITY
Basic idea of probability 1. Suppose you are selecting a single marble from a jar that contains 50 black and 50 white marbles. ( In this example, the jar of marbles is the population, and the single marble is the sample). Although we cannot guarantee the exact outcome of our sample, it is possible to talk about the potential outcomes in term of probabilities. In this case we have 50-50 chance of getting either color. 2. Consider another jar that has 90 black and 10 white marbles. Again we cannot specify the exact outcome of the sample, but now we know that the sample probability will be a black marble.

By knowing the make up of a population, we can determine the probability of obtaining specific samples. In this way, probability gives us a connection between populations and samples, and this connection will be the foundation for the inferential statistics.
3

THE ROLE OF PROBABILITY IN INFERENTIAL STATISTICS

In the preceding examples we begin with a population and then use probability to describe the samples that could be obtained. This is exactly backward from what we want to do with inferential statistics where we begin with a sample and then answer general question about the population. To do that, first we develop probability as a bridge from populations to samples. This stage involves identifying the types of samples that probably would be obtained from a specific population. Once this bridge is established, we will reverse the probability rules to allow us to move from samples to population Figure 6.1).
4

THE ROLE OF PROBABILITY IN INFERENTIAL STATISTICS

Insert fig 6.1 p163 G

WHAT IS PROBABILITY
For a situation in which several different outcome are possible, the probability for any specific outcome is defined as a fraction or a proportion of all the possible outcomes. If the possible outcome are identified as A,B,C,D and so on, then: Probability of A =
number of outcomes classified as A total number of possible outcomes

The probability of a specific outcome will be express with a p (for probability) followed by the specific outcome in parentheses. For example, the probability of obtaining heads for a coin toss will be written as p (heads).
6

RANDOM SAMPLING
A random sample requires that each individual in the population has an equal chance of being selected. A second requirement, necessary for many statistical formulas, states that the probabilities must stay constant from one selection to the next if more than one individual is selected.

PROBABILITY AND FREQUENCY DISTRIBUTIONS

The situation in which we are concerned with probability usually will involve a population of scores that can be displayed in a frequency distribution graph. If you think of the graph as representing the entire population, then different portions of the graph will represent different portions of the population. Because probabilities and proportions are equivalent, a particular portion of the graph corresponds to a particular probability in the population. Thus, whenever a population is presented in a frequency distribution graph, it will be possible to represent probabilities as portion of the graph.
8

PROBABILITY AND FREQUENCY DISTRIBUTIONS

Example. A simple population that contains only N=10 scores with values 1,1,3,3,4,4,4,5,6. This population is shown in the frequency distribution graph in Figure 6.2. What is the probability of obtaining an individual with a score greater than 4? p(X>4) = ? The answer is the shaded part of the distribution that is 2 squares out of the total 10 squares in the distribution.
Notice that we are now defining probability as a proportion of area in the frequency distribution graph.
9

PROBABILITY AND FREQUENCY DISTRIBUTIONS

10

PROBABILITY AND THE NORMAL DISTRIBUTION

Now that you have better understanding of the concept of probability, and we are learning how to use z-score to determine probabilities. Learning how to determine probability using the normal distribution will help us master other skills required in the future. Example Assume that the population of adult heights forms a normalshaped distribution with a mean = 68 inches and a standard deviation of = 6 inches. Given this information (see Figure 6.4), we can determine the probabilities associated with specific samples. For example, what is the probability of randomly selecting an individual from this population who is taller than 80 inches?
11

PROBABILITY AND THE NORMAL DISTRIBUTION

12

PROBABILITY AND THE NORMAL DISTRIBUTION

Restating this question in probability notation, we get p(X>80) = ? Following are the steps to find the answer: 1. Translate the probability question into a proportion question. Out of all possible adult heights, what proportion is greater than 80 inches? 2.The set of all possible adult heights is simply the population distribution. Because we are interested in all heights greater than 80, we shade in the area to the right of 80. This area represents the proportion we are trying to determine. 3. Identify the exact position of X=80 by computing z-score.

z= = = 68= 2.012 80 X 4. The proportion we are trying to determine 6 6may now be express in term of its z-score, i.e. p(z>2.0) = ? According to the proportion shown in the normal distribution graph, we have 2.15% of the scores in the tail beyond z = +2.00. So the answer is 2.15%.
13

NORMAL DISTRIBUTION AND AREA

Insert fig 6.2 p 120 R

14

THE UNIT NORMAL TABLE

The graph of standard normal distribution shows proportions for only a few selected z-score values. A more complete listing of zscores and proportions is provided in the unit normal table. This table lists proportion of the normal distribution for a full range of possible z-score values. The vertical lines separate the distribution into two section, the body and the tail. Part of the table is reproduced in Figure 6.6. Notice that the table is structured in a four-column format. The z-score values listed in column A. Column B presents the body, column C presents the portion in the tail, and column D identifies the proportion of the distribution between the mean and the z-score.
15

THE UNIT NORMAL TABLE

16

PROBABILITY,PROPORTIONS, AND zSCORES

The unit normal table lists relationships between z-score locations and proportions in a normal distribution. For any z-score location, we can use the table to look up the corresponding proportion. Because we have defined probability as equivalent to proportion, we can also use the unit normal table to look up probabilities for normal distribution.

17

PROBABILITY,PROPORTIONS, AND zSCORES

The unit normal table lists relationships between z-score locations and proportions in a normal distribution. For any z-score location, we can use the table to look up the corresponding proportions, and since we have defined probability as equivalent to proportion, we can also use the unit normal table to look up probabilities for normal distribution.
18

PROBABILITY,PROPORTIONS, AND zSCORES

Insert fig 6.7 p172

19

FINDING PROPORTIONS/PROBABILITIES FOR SPECIFIC zSCORE VALUES

Example A What proportion of the normal distribution corresponds to zscore values greater than Z = 1.00? Solution: 1. Sketch the distribution and shade the area you are trying to determine (refer fo Figure 6.7A). 2. Look up z=1.00 in column A of the unit normal table, than read column C (tail) for the proportion. We find the answer is 0.1587. 3. This problem could be rephrased as What is the probability of selecting a z-score value greater than z = 1.00+ ? The answer is p(z>1.00) = 0.1587 (or 15.87%)
20

FINDING PROPORTIONS/PROBABILITIES FOR SPECIFIC zSCORE VALUES EXAMPLE B For normal distribution, what is the probability of selecting a zscore less than z = 1.50 (In symbol p(z<1.50 = ?) Solution: 1. Sketch the distribution and shade the area you are trying to determine (refer to Figure 6.7B). Note that we have shaded all the values to the left of z = 1.5 and this is the portion we are trying to find. 2. Find z = 1.50 in the unit normal table and read the proportion from column B. The answer is 0.9332. The answer is p(z<1.50) = 0.9332 (or 93.32%).
21

FINDING PROPORTIONS/PROBABILITIES FOR SPECIFIC zSCORE VALUES

Example C Many problems will require that we find proportions for negative zscores. For example, what proportion of the normal distribution corresponds to the tail beyond z =-1.50? That is p(z<-0.50) = ? Solution: (To answer questions with negative z-scores, simply remember that the normal distribution is symmetrical. The proportion in the left tail is identical to the proportion in the right tail.) 1. Sketch the distribution and shade the area you are trying to determine (refer to Figure 6.7C). 2. Find z = -0.50 in the unit normal table and read the proportion from column C. The answer is 0.3085. The answer is p(z<-0.50) = 0.3085 (or 30. 85%).
22

PROBABILITIES AND PROPORTIONS FOR SCORES FROM A NORMAL DISTRIBUTION

In the preceding examples, we used the unit normal table to find probabilities and proportions corresponding to specific z-score values. In most situations, however, it will be necessary to find probabilities for specific X values. Example It is known that IQ scores from a normal distribution with = 100 and =15. Given this information, what is the probability of randomly selecting an individual with an IQ score less than 130? (p(X<130) = ?. Solution: Step 1. Transform the X values into z-scores z=
X

Step 2. Use the unit normal table to look up the proportions corresponding to the z-score values. (p(X<130) = 0.9772 (or 97.72%)
23

130 = 100 15

= 2.00 30
15

PROBABILITIES AND PROPORTIONS FOR SCORES FROM A NORMAL DISTRIBUTION

Insert fig 6.9Gp176

24

EXAMPLE The highway department conducted a study measuring driving speeds on a local highway. They found an average speed of =58 km per hour with a standard deviation of =10. The distribution is approximately normal. Given this information, what proportion of the cars are traveling between 55 and 65 km per hour? Step 1. Determine the z=score corresponding to the X value at each end of the interval.

PROBABILITIES AND PROPORTIONS FOR SCORES FROM A NORMAL DISTRIBUTION

For X=55; z =
For X=65; z =

X =

= X = 65 58 7 seeking can 10 be divided 10 The proportion we are into two section; the area on the left of

= 58 55 10

= -0.30 3 10 = 0.70

the mean and the area on the right of the mean. Using column D of the unit normal table

p(55 < X < 65) = p(-0.30 < z < +0.70) = 0.1179 + 0.2580 = 0.3759 (or 37.59%)
25

PROBABILITIES AND PROPORTIONS FOR SCORES FROM A NORMAL DISTRIBUTION

Insert fig6.10Gp177

26

PROBABILITIES, PROPORTIONS, AND PERCENTILE RANKS

Thus far we have discussed parts of distribution in term of proportions and probabilities. However, there is another set of terminology that deals with many of the same concepts, i.e. percentile ranks. As we have defined, the percentile rank for a specific score as the percentage of the individuals in the distribution who have scores that are less than or equal to the specific score. Using this terminology, it is possible to rephrase some of the probability problems, for example, the problem What is the probability of randomly selecting an individual with IQ of less than 130? could be rephrased as What is the percentile rank for IQ score of 130?
27

PROBABILITY AND INFERENTIAL STATISTICS

Probability forms a direct link between samples and populations from which they come. This link will be the foundation for the inferential statistics. The following example provides a brief preview of how probability will be used in the context of inferential statistics. Example A research begins with a population that forms a normal distribution with a mean of = 400 and a standard deviation = 20. A sample is selected from the population and a treatment is administered to the sample (Figure 6.19). The goal for the study is to evaluate the effect of the treatment. To determine whether or not the treatment has an effect, the researcher simply compares the treated sample with the original population. If the individual in the sample have scores around 400, then we must conclude that the treatment appears to have no effect. On the other hand, if the treated individuals have scores that are noticeably different from 400, then the researcher has evidence that the treatment does have an effect. The problem for the researcher is determining exactly what is mean by noticeably different from 400.
28

PROBABILITY AND INFERENTIAL STATISTICS

Insert fig6.19 g p187

29

PROBABILITY AND INFERENTIAL STATISTICS

Notice that the study is using a sample to help answer a question about a population; this is the essence of inferential statistics. Figure 6.20 shows the original population from our hypothetical research study. We use z-scores to resolve this problem and select z-score value beyond z=+1.96 (or z=-1.96) as an extreme value and therefore noticeable different. The boundaries provide objective criteria for deciding whether or not our sample is noticeable different from the original population. Specifically, a sample that falls in this extreme area is extremely unlikely and is defined as a probability that is 5% or less. Note that probabilities allow us to separate a distribution into those scores that are likely to be obtained (high probability) and those scores that are extremely unlikely (low probability). So if our treated individual has a score that is located in the extreme 5% area, we can reach a logical conclusion that the treatment had an effect.

30

PROBABILITY AND INFERENTIAL STATISTICS

Insert fig 6.20 p188G

31