Presentation is loading. Please wait.

Presentation is loading. Please wait.

The Process of Data Analysis February 14, 2011. 2 Objectives By the end of this meeting, participants should be able to: a)Describe the various ways in.

Similar presentations


Presentation on theme: "The Process of Data Analysis February 14, 2011. 2 Objectives By the end of this meeting, participants should be able to: a)Describe the various ways in."— Presentation transcript:

1 The Process of Data Analysis February 14, 2011

2 2 Objectives By the end of this meeting, participants should be able to: a)Describe the various ways in which causality can occur and draw causal diagrams. b)Develop a hypothesis based on a causal theory and explain the necessary considerations for testing a hypothesis. c)Distinguish variables by their level of measurement. d)Define Type I and Type II error and explain why they should be avoided.

3 3 How to Start Data Analysis a)Decide whether you are interested in a descriptive or explanatory analysis b)If you decide on the latter, state at least one hypothesis (could be several) c)Carefully identify your variables or predictors Independent (a cause or a predictor) Independent (a cause or a predictor) Dependent (a result) Dependent (a result)

4 4 Thinking about Causation a)Causation can occur in many ways Direct Direct Indirect Indirect Multiple Multiple Reciprocal Reciprocal b)Due to the fact that the most interesting phenomenon frequently have different types of causes, researchers frequently draw causal diagrams

5 5 Testing a Causal Hypothesis Four tests are necessary for an independent variable to cause a change in a dependent variable a)Association b)Temporal order c)Need to consider alternative causes d)Causal Mechanism South Korean attitudes towards contraceptives and appliance ownership South Korean attitudes towards contraceptives and appliance ownership Homicide rate and ice cream sales Homicide rate and ice cream sales

6 6 Operationalizing Concepts a)Researchers must find a way to change concepts that they seek to measure into tangible variables b)For example, if you believed that wealth impacted vote choice, how should that concept be operationalized?

7 7 Levels of Measurement a) a)Nominal Variables: Distinct categories that are not related in any numerical or orderly fashion. (No order) Ex: respondents ’ religion, region of the country, or race b) b)Ordinal Variables: Ordered categories that do not have any intrinsic numerical qualities. (Ordered but uneven intervals) Ex: SA, A, N, D, SD or 1st, 2nd, 3rd, 4 th Party?

8 8 Levels of Measurement a) a)Interval Variables: Intrinsically numeric but lack a meaningful zero point; distance between successive numbers are equal. (Ordered and even intervals) Ex: Fahrenheit scale or Celsius scale b) b)Ratio Variables: Intrinsically numeric with a meaningful zero point. (Ordered, even intervals, and a meaningful zero point) Ex: Age, height, and income

9 9 Hypothesis Testing a) a)Hypothesis (H): Eating ice cream builds muscle mass. b) b)Fallacy of affirming the consequent: I am stronger; it must be from eating ice cream. c) c)Denying the consequent: If I am not stronger, I couldn ’ t have eaten any ice cream. d) d)A research hypothesis cannot be proven, only disproven. e) e)If I eat more ice cream and I get stronger, it does not mean that eating ice cream builds muscle mass.

10 10 Hypothesis Testing a) a)Null Hypothesis (Ho): Opposite of the research hypothesis. In this case, no relationship exists between eating ice cream and building muscle. b) b)My hypothesis predicts that we will reject the null hypothesis. c) c)Rejecting the null hypothesis does not prove the research hypothesis. Rather, by offering evidence that the null is not true, the researcher shows that the research hypothesis may be true.

11 11 Types of Error a) a)Type 1 Error: Reject (incorrectly) a true null hypothesis. The null hypothesis (that no relationship exists) is true. Our analysis, however, incorrectly leads us to conclude that a relationship exists. b) b)Type 2 Error: Accept (incorrectly) a false null hypothesis. The null hypothesis (that no relationship exists) is false. Our analysis, however, incorrectly leads us to conclude that no relationship exists.

12 12 Statistical Inference a)Researchers need to how much chance of being incorrect are they willing take. b)The standard acceptable Type I Error in the social sciences is.05.

13 13 For Next Time a)Read WKB chapter 9. b)Open the class poll from j.mp/pubOpin c)Using the variables from the poll, derive three distinct hypotheses and identify what level of measurement the variables are (nominal, ordinal, interval, ratio) Hypotheses need to clearly state a possible relationship between variables in the class survey Hypotheses need to clearly state a possible relationship between variables in the class survey Levels of measurement need to reflect the questions as asked on the survey (not how they could have been asked). Levels of measurement need to reflect the questions as asked on the survey (not how they could have been asked). Think about a causal mechanism. (Don ’ t have to write that out right now.) Think about a causal mechanism. (Don ’ t have to write that out right now.)


Download ppt "The Process of Data Analysis February 14, 2011. 2 Objectives By the end of this meeting, participants should be able to: a)Describe the various ways in."

Similar presentations


Ads by Google