In order to draw relevant conclusions from any given data set, the researcher needs to address the following points: There must be clear delineation between continuous and discrete variables. Continuous variables have values that can be measured in intervals. Discrete variables, on the other hand, have values that cannot be measured in intervals (in whole numbers or integers). Examples of continuous variables are temperature, weight, income, height, and depth. Examples of discrete variables are age, number of months, and number of children;
The second consideration is related to the type of statistical tests used in specific analysis. The Student’s t-test is used for comparing the variance of ‘two’ sample groups. The ANOVA or the Analysis of Variance is used for more than two sample groups (it can also be used in analyzing the variance of two sample groups. Regression analysis provides a synthetic equation which describes the relationship between one or more variables; In group analysis, correlation is of little value.
The important thing is the perpetuity of the variance or the existence of disparity; It would be irresponsible for the researcher to conduct multiple statistical tests when only one or two is required to establish the validity of any given hypothesis; The researcher should always consider the normality of any given data set. If the data set is normally distributed, he/she may use the ANOVA or the t-test. If, however, the data set is not normally distributed, he/she should use nonparametric tests like the Kruskal-Wallis Test.
Lastly, the researcher must maintain internal validity and reliability in his/her methodologies. This is important because research is grounded on validity and reliability. Manipulation or undue ‘dropping’ of variables is always the ‘mortal sin’ of quantitative research. In the data set provided, note that continuous and discrete variables are not grouped. The researcher should therefore provide a necessary format for an easier analysis of data. Items with blank ‘content’ are automatically assigned a ‘0’ value.
If the associated variable is an ordinal one, retaining the blank item is acceptable. If, however, the associated variable is a nominal one, then the researcher should remove that sample from the data set. Retaining ‘empty’ values in rank analysis often leads to confusion and misinterpretation of data. The size of any sample always follows from the size of the universe or the population. If the population is about 120 000, then the sample size should be around 4 to 5 percent. If the population is more or less 1000, then the sample size should be around 40 to 100.
Note that if the sample size is equal or less than 25, then the researcher should use the Student’s t-test. Suppose the sample size is greater than 25 (congruent to the population), the researcher may use the Z-test (assuming that the data is normally distributed). Now, the so-called p-value should not be confused with the P(X) – the former indicates the critical area in a distribution while the latter the probability of any given event. Finally, data analysis is insufficient if not supplemented by discourses on related literature, theories, and hypotheses.
Hypothesis Testing Hypothesis I: The age of residents of region I is greater than the age of residents of region VII. Null Hypothesis: There is no difference in the age of residents of region I and the age of residents of region VII. Alternative Hypothesis: The age of residents of region I is greater than the age of residents of region VII. Preliminary Analysis The variables ‘age’ and ‘region’ are discrete variables. It is better to plot discrete variables in bar graphs than in line graphs. ANOVA may be used in the analysis. Note that analysis is one-tailed.