Using Statistics in HRD
Martelli, Joseph T., Training & Development
For many HRD professionals, using statistics seems a daunting task - but it needn't be. For some, the fear of statistics conjures a former college course on the subject, and the thought of memorizing all those formulas and performing tedious calculations again can be mindboggling.
This article may help change the way you think about and use statistics. Statistics are not something to be intimidated by or fearful of. In fact, using them can make your job easier. Today, the last thing you should be doing is calculating statistics by hand. Instead, let a PC-based software package do the calculating for you. PC-based statistical packages are abundant, and the prices are affordable.
While you can let your PC do the numbers crunching for you, it is necessary to understand how to interpret and use statistics appropriately; no PC can do that. In order to properly use statistics, there are some important concepts and principles you must understand.
Classification of statistics
Statistics come in two types: descriptive and inferential. As the name implies, descriptive statistics describe data. Three of the most common descriptive statistics are called measures of central tendency: the mean, the median, and the mode. These are what most non-statisticians call the "average" and, as we will see below, each serves a purpose.
Most HRD professionals use descriptive statistics regularly, whether they know it or not. Descriptive statistics are frequently used in conjunction with charts and graphs to display data visually. Charts and graphs are an easy and simple way to visualize trends that have large amounts of data.
The more powerful types of statistics are inferential statistics. Inferential statistics let the user make inferences about a population using a sample drawn from the population. With inferential statistics (and a solid research design), the user can estimate the probability that a relationship observed in the sample is also true for the larger population. With descriptives, you are limited primarily to describing associations within the sample itself.
Classification of data
HRD professionals are constantly collecting data, whether they know it or not. Evaluation sheets from training programs, questionnaires and surveys, training program cost data, program development costs, and needs analysis results, all represent data. Being able to handle or process the data is what discriminates the novice from the seasoned professional.
It's important to understand the levels of data, and to be aware of the level of data for each piece that you work with. For example, a questionnaire, even a short two-pager, may have up to 100 measures, and each data point has its own level.
There are four primary classifications of data.
* Nominal. Nominal data are the simplest types of data we have. Mathematical operations (+, -, x, and/) make no sense for nominal-level data. Nominal-level data, from a statistical perspective, lets us code or classify items into groups. Examples of nominal data are an employee's sex, department, or occupation. Their purpose is merely to classify or identify people or things within a larger system. The mode (the most frequent item) as a measure of central tendency is appropriate for use with nominal-level data.
* Ordinal. Ordinal-level data have all the characteristics of nominal data, but, in addition, these data imply a rank order of importance. Take for instance a performance appraisal that simply rates whether an employee's performance is poor, fair, good, or excellent. With this ordinal measure, you can rank people according to their performance, but you can't tell, for example, the difference between "good" performance and "fair" performance. In addition to the mode, you can calculate the median when using ordinal data. The median is the value that falls in the middle when all the data points are ranked.
* Interval. …