PSYCHOLOGICAL RESEARCH (PYC 304-C) Lecture 2 1. Continuous variables (continued from Lecture 1) 1.1 Ratio-scale variables The main characteristic of a ratio-scale variable is that it has an absolute zero, that is, a well-defined starting point of reference. For example, a height of 0 centimetres is universally and easily understood. Hence, the ratio of two heights will always be the same, irrespective of the units that we use. If a person is twice as tall as another, he will remain so if height is measured in centimetres, metres, feet or inches. 1.2 Interval-scale variables As opposed to a ratio-scale variable, this type of variable does not have a typical absolute zero on its scale. The variable temperature is a very good example when benchmarked, its zeros on the Celsius, Fahrenheit and Kelvin scales prove to be completely different from one another! In fact, we should know that 0 o C = 32 o F = 273 o K. In real-life, if the temperature in Durban is 30 o C and that in Cape Town is 15 o C, then we cannot say that it is twice as hot in Durban than it is in Cape Town, the simple reason being that, if we were to convert these temperatures to degrees Fahrenheit, they would be 86 o F and 59 o F respectively in Durban and Cape Town (now, 86 is not twice 59!). In the case of interval-scale variables, we may only say that the difference between the two temperatures is 15 o C. 2. Parameters and statistics Bearing in mind the definitions of sample and population given in the Lecture 1, we now try to explain how the concept of estimation rests on the terms parameters and statistics. A parameter is a characteristic of a population the mean, variance and proportion are three examples. If it were possible to carry out a census, then these values would be calculated exactly. Parameters are actual or real values but seldom can they be determined since, for various practical reasons, it is impossible to carry out a census. Information therefore has to be gathered by means of a sample that should be selected in order to be as representative as possible of its parent population (by using the most appropriate sampling method.) 1
Should a sample be a perfect replica of the population, it would be unbiased, that is, completely free of sampling errors (bias) but then again, it is practically impossible to select an ideal sample! We can thus only estimate population parameters on the basis of data obtained from a sample. The sample mean, variance and proportion are examples of statistics. Population Parameters: mean µ 2 variance σ proportion p Sample Statistics: mean x variance proportion 2 s p s Fig. 2.1 Parameters and statistics From the above discussion, we may deduce the following (see Fig. 2.1): 1. Parameters are actual population characteristics. 2. Statistics are calculated from samples. 3. Statistics are estimators of population parameters. For example, the sample mean x is an estimator of the population mean µ. 3. Scientific research Scientific research is carried out by using empirical methods, namely quantitative techniques. In pure and industrial psychology, we often come across constructs that can hardly be quantified, because of their intrinsically qualitative nature, without losing the flavour of their meaning. However, for the purpose of scientific investigation, statistical analysis and hypothesis testing dictate us to assign numerical values to these variables and their attributes (as will be seen later). Quantitative techniques may be summarized as the following: 1. Collection of data 2. Presentation of data 3. Analysis and interpretation of data 4. Decision-making and/or forecasting 2
3.1 Knowledge acquisition There are various sources, with different degrees of reliability, from which to collect relevant data: 1. Superstition and custom 2. Intuition 3. Respected sources 4. Reasoning 5. Experience 6. Scientific investigation In this module, we will specifically focus on scientific investigation. 3.2 Characteristics of scientific research 3.2.1 Maintaining control In scientific experimentation, the researcher must be able to control independent (explanatory) variables, or predictors, to a certain extent, if not totally, in order to observe their effects on the dependent (response) variable. As an example, if we were to heat a metal rod and observe the effect on its physical appearance, the dependent variable would probably be length of rod and the independent variable, temperature. We thus manipulate temperature so as to record the change in length. Most experiments involve more than one independent variable, a typical example being the effect of a mixture of qualitative and quantitative variables on exam marks (obtained by students). We can think of 1. Attendance rate 2. Number of hours spent studying 3. Number of hours spent sleeping 4. Anxiety 5. Intelligence 6. Number of hours spent watching television as predictors, each having its own strength level. It is worthwhile noting that constructs like anxiety and intelligence cannot be manipulated by the researcher whilst the variable attendance rate is not only easily quantifiable but can be imposed on students (that is, manipulated) as well. 3
3.2.2 Operationalisation of definitions In psychological research, there are constructs or variables that simply cannot be measured because of their highly abstract and qualitative nature. An interesting example is anxiety the researcher has to operationalise its definition in order to show how it is going to be measured. In fact, anxiety is a concept created by scientists from observable behaviours like 1. Sweat on forehead and hands 2. Empty sensation in the stomach 3. Heart palpitations which are known as behavioural instances and statistically termed as manifest variables. Thus, anxiety is measurable through its manifest variables which themselves require the development of an appropriate measurement scale (to be discussed later). 3.2.3 Repeatability Repeatability is synonymous to consistency one should be able to replicate any scientific investigation, irrespective of the conditions and circumstances. After all, research results are meant to be shared with the community so as to consolidate or freshen up existing theory. The well-known concept of reliability takes care of the repeatability property of research. The investigator must ensure that his/her research is, above all, reliable in terms of data collection and research methodology. The measuring instrument (e.g. questionnaire) must pass reliability tests, nowadays easily carried out by a statistical software like SPSS, before results are published. 3.3 Types of research In this module, we will only consider two types of researches: 1. Descriptive 2. Experimental TYPES OF RESEARCH Descriptive No attempt to establish any cause-effect relationships Experimental Seeks causal relationships between variables Fig. 3.1 Types of research 4
3.3.1 Descriptive research The objectives of descriptive research are collection of data and their presentation in the forms of diagrams, charts and graphs. There is no attempt to deduce any causal relationships between variables or constructs (see Fig. 3.1). Data may be retrieved from primary (survey and experiments) as well as secondary sources (past records and archives). Several types of investigations may be carried out as illustrated in Fig. 3.2 below. N = 1 Self-reporting Case study Naturalistic observation Correlational study Ex post facto study DESCRIPTIVE RESEARCH Surveys and interviews Cross-sectional study Longitudinal study Meta-analysis Participant observation Fig. 3.2 Descriptive research A brief description of each type of study is given in Fig. 3.3 below. 5
Self-reporting Introspective thinking that qualifies as a scientific approach if it is used as a potential source of ideas and conjectures to decide whether hypotheses are acceptable. Correlational study Describes the degree of relationship between two variables and makes it possible to predict the behaviour of one variable given information on the other. Does not allow making causal inferences. Case study Intensive description and analysis of an individual, organisation or event. Reports are more subjective, less data-oriented and more clinically oriented. Longitudinal study Exploration of evolutionary changes over time by measuring the same attributes of the same group of people at regular intervals. Cross-sectional study Exploration of evolutionary changes over time by measuring the same attributes in representative samples of individuals at different construct levels. Surveys and interviews Collection of data on participants in a standardised form (e.g. questionnaire) from a representative sample of the population. The measuring instrument must be valid and reliable to allow for making inferences. N = 1 Study involving only one individual as well as a very clearly identifiable dependent variable. Useful for studying the effects of different types of therapy on certain behavioural problems. Not used to test theory. DESCRIPTIVE RESEARCH COLLECTION OF DATA AND TYPES OF ANALYSIS Participant observation Observation and recording of the behaviour of others by active participation of researcher, that is, he/she forms part and interacts with individuals from the target population. Meta-analysis Quantitative technique that integrates and describes the results of a large number of studies carried out by the same person. After literature search, statistical analyses are applied in order to answer a research question. Naturalistic observation Unobtrusive collection of data on spontaneous behaviour (candid-camera type). Fig. 3.3 Collection of data and types of analysis Ex post facto study Comparison of the effects of two or more variables with the exception that these variables cannot be manipulated (e.g. gender and age). 6