Some Introductory Statistics Terminology

Descriptive Statistics Procedures used to summarize, organize, and simplify data (data being a collection of measurements or observations) taken from a sample Examples: –Expressed on a 1 to 5 scale, the average satisfaction score was 3.7 –43% of students in an online course cited that family obligations were the main motivation behind choosing distance education

Inferential Statistics Techniques that allow us to make inferences about a population based on data that we gather from a sample Study results will vary from sample to sample strictly due to random chance (i.e., sampling error) Inferential statistics allow us to determine how likely it is to obtain a set of results from a single sample This is also known as testing for “statistical significance”

Population A population is the entire set of individuals that we are interested in studying This is the group that we want to generalize, or apply, our results to Although populations can vary in size, they are usually quite large Thus, it is usually not feasible to collect data from the entire population

Sample A sample is simply a subset of individuals selected from the population In the best case, the sample will be representative of the population That is, the characteristics of the individuals in the sample will mirror those in the population

Variables A characteristic that takes on different values for different individuals in a sample Examples: –Gender –Age –Course satisfaction –The amount of instructor contact during the semester

Independent Variables (IV) The “explanatory” variable The variable that attempts to explain or is purported to cause differences in a second variable Example: –Does the use of a computer-delivered curriculum enhance student achievement? –Whether or not (yes or no) students received the computer instruction is the IV

Dependent Variables (DV) The “outcome” variable The variable that is thought to be influenced by the independent variable Example: –Does the use of a computer-delivered curriculum enhance student achievement? –Student achievement is the DV

Confounding Variables Researchers are usually only interested in the relationship between the IV and DV Confounding variables represent unwanted sources of influence on the DV, and are sometimes referred to as “nuisance” variables Example: –Does the use of a computer-delivered curriculum enhance student achievement? –One’s previous experience with computers, age, gender, SES, etc. may all be confounding variables

Controlling Confounding Variables Typically, researchers are interested in excluding, or controlling for, the effects of confounding variables This is not a statistical issue, but is accomplished by the research design Certain types of designs (e.g., true experiments) better control the effects of confounding variables

Central Tendency

Measures of Central Tendency Three measures of central tendency are available –The Mean –The Median –The Mode Unfortunately, no single measure of central tendency works best in all circumstances –Nor will they necessarily give you the same answer

Example SAT scores from a sample of 10 college applicants yielded the following: –Mode:480 –Median: 505 –Mean: 526 Which measure of central tendency is most appropriate?

The Mean The mean is simply the arithmetic average The mean would be the amount that each individual would get if we took the total and divided it up equally among everyone in the sample Alternatively, the mean can be viewed as the balancing point in the distribution of scores (i.e., the distances for the scores above and below the mean cancel out)

The Median The median is the score that splits the distribution exactly in half 50% of the scores fall above the median and 50% fall below The median is also known as the 50th percentile, because it is the score at which 50% of the people fall below

Special Notes A desirable characteristic of the median is that it is not affected by extreme scores Example: –Sample 1: 18, 19, 20, 22, 24 –Sample 2: 18, 19, 20, 22, 47 –The median is 20 in both samples Thus, the median is not distorted by skewed distributions

The Mode The mode is simply the most common score There is no formula for the mode When using a frequency distribution, the mode is simply the score (or interval) that has the highest frequency value When using a histogram, the mode is the score (or interval) that corresponds to the tallest bar

Choosing the Proper Statistic Continuous data –Always report the mean –If data are substantially skewed, it is appropriate to use the median as well Categorical data –For nominal data you can only use the mode –For ordinal data the median is appropriate (although people often use the mean)

Distribution Shape and Central Tendency In a normal distribution, the mean, median, and mode will be approximately equal

Distribution Shape (2) In a skewed distribution, the mode will be the peak, the mean will be pulled toward the tail, and the median will fall in the middle

Frequency Distribution Tables

Overview After collecting data, researchers are faced with pages of unorganized numbers, stacks of survey responses, etc. The goal of descriptive statistics is to aggregate the individual scores (datum) in a way that can be readily summarized A frequency distribution table can be used to get “picture” of how scores were distributed

Frequency Distributions A frequency distribution displays the number (or percent) of individuals that obtained a particular score or fell in a particular category As such, these tables provide a picture of where people respond across the range of the measurement scale One goal is to determine where the majority of respondents were located

When To Use Frequency Tables Frequency distributions and tables can be used to answer all descriptive research questions It is important to always examine frequency distributions on the IV and DV when answering comparative and relationship questions

Three Components of a Frequency Distribution Table Frequency –the number of individuals that obtained a particular score (or response) Percent –The corresponding percentage of individuals that obtained a particular score Cumulative Percent –The percentage of individuals that fell at or below a particular score (not relevant for nominal variables)

Example (1) Frequency distribution showing the ages of students who took the online course

Example (2) Student responses when asked whether or not they would recommend the online course to others Most would recommend the course

Independent t-Test

The independent samples t-test is used to test comparative research questions That is, it tests for differences in two group means –Two groups are compared on a continuous DV

Scenario Suppose we wish to compare how males and females differed with respect to their satisfaction with an online course The null hypothesis states that men and women have identical levels of satisfaction

Research Question If we were conducting this study, the research question could be written as follows: –Are there differences between males and females with respect to satisfaction? The word “differences” was used to denote a comparative question

The Data (1) Satisfaction is measured on a 25-point scale that ranges between 5 (low) and 30 (high) The descriptive statistics were as follows:

The Data (2) On a 25-point satisfaction scale, men and women differed by about 5 points (means were and 23.5, respectively) They were not identical, but how likely is a 5 point difference to occur from the hypothetical population where men and women are identical?

Conceptual Formula The conceptual formula for the t statistic is The formula tells how big the 5 point difference we observed is relative to the difference expected simply due to sampling error

Results The t-statistic value was 1.695, suggesting that the 5-point difference is not quite twice as large as the difference we would expect due to chance (which is quantified by the standard error statistic) The p-value for the analysis was.116 (almost.12, or 12%)

Interpreting the Probability Thus, there was about a 12% chance that this sample (the 5 point difference) originated from the hypothetical null hypothesis population The p-value is greater than.05, so we would retain the null (results are not significant) Thus, there is no evidence that males and females differ in their satisfaction

Cohen’s d Effect Size Recall that p-values don’t tell how important the results are A measure of effect size can be computed that helps us quantify the magnitude of the results we obtained The mean difference (5 points) is expressed in standard deviation units

Example Using the statistics from the SPSS printout, the d effect size can be computed as

Interpreting Cohen’s d Cohen (1988) suggested the following guidelines for interpreting the d effect size –d >.20 is a small effect size (1/5 of a standard deviation difference) –d >.50 is a medium effect size (1/2 of a standard deviation difference) –d >.80 is a large effect size (4/5 of a standard deviation difference)

Writing Up the Results If you were writing the results for publication, it could go something like this: –“As seen in Table 1, satisfaction scores for female students were approximately five points higher, on average, than those of males. Using an independent t test, no statistically significant differences were observed between the group means, (t (12) = 1.70, p =.12). However, despite no statistical significance, Cohen’s d effect size indicated a large difference between the groups (d =.92)”