# Learning statistics, cannot identify what a study is measuring

I have to describe the statistics of a study. I chose a study attempting to improve detection of mentally ill detainees by adding an additional questionnaire. The officers use the PISP and the study added the RDS questionnaire interview to see if this improved screening for mentally ill inmates coming through a small jail.

Most questions are yes or no responses on the PISP, and the RDS is done by student-interview and then the SADS-C was done by clinician interview with detainees.

They provide validity, convergence, etc. values. My question is what type of statistics is this? What is really being measured? Is this qualitative or quantitative, descriptive or inferential, is the data nominal-categorical? Can someone help me see the basics of this type of study using vocab?

You seem to have a number of questions regarding the study. I'll try to give you a quick primer on some of the vocabulary, then you can figure out which vocab is relevant to the study. If you want your own reference, any statistics book will have discussion of vocabulary in the first chapter (or maybe second). There are also a number of online resources for reviewing stats vocab. Here are two, there are plenty of youtube videos, or you can check out a resource like Khan Academy for other instructional videos

• Descriptive statistics: These are statistics that are just describe a particular sample. Average height, average weight, education level, etc. would all be descriptive statistics.
• Inferential statistics: This is when you are actually trying to make an inference about a sample (i.e. answering a question about it). An example of a question you might be trying to answer is, "Do taller soccer teams win more?" In something like this, you would use descriptive statistics to make an inference through a statistical test.

Now let's talk about Qualitative vs. Quantitative measures.

• Qualitative measures: These are measurements you make about a particular quality/category that a sample falls into. Examples would be male/female, eye color, state of residence.
• Quantitative measures: These are measurements of a particular quantity. Examples would be money spent on groceries, last test score, number of goals scored.

For type of data, you can have a couple different kinds. Qualitative Data can be either Nominal or Ordinal. Quantitative data can be either discrete or continuous.

• Nomimal data: This is categorical data, where people are placed in different categories that have no obvious relationship between them. Male/female is a classic one. Male isn't necessary better or worse than female, they're just different categories.
• Ordinal data: This is similar to nominal data, but there's a relationship between the different values. However, the relationship isn't necessarily standard. An example are Likert items (questions where the answers are "Strongly disagree/Disagree/Neutral/Agree/Strongly Agree"). In a Likert question, answering "Strongly Agree" is clearly a stronger response than "Agree", but you can't say that the difference between "Neutral" and "Agree" is the same as the difference between "Agree" and "Strongly Agree".
• Discrete data: Data that takes on specific values. "How many cars do you own?" is an example. You can own 0, 1, 2, 3, etc. cars. You can't own 2.5 cars.
• Continuous data: This is data that has a continuous spectrum. It can literally take on any value (though often we are limited by our measuring tools). An example of this would be the amount of rain at a particular location. This can take on any value from 0 to infinity. We could have 2 inches of rain, 2.5, 2.55, 2.551, etc.

So there's a basic primer on the vocabulary you mentioned. The resources linked above talk about some other terms that are important in statistics but I think this covers the basics of what you'll need to answer your questions.

• Very helpful thanks so much for the resources, too! – Gail Amalfitano Apr 1 '15 at 5:35
• Now we are discussing power and effect.I understand effect size and Cohen's d is one method to measure effect size.I struggled with understanding the power,type I and II errors,rejecting or accepting the null, and how these all relate to power and effect size. Why is an effect less meaningful if the sample is too large like on SAT or ACT tests males perform better in math than females, but the effect size is negligible due to the very large sample size..why? We are learning how to appraise research and studies.After this i must distinguish qualitative from quantitative studies ...any advice? – Gail Amalfitano Apr 1 '15 at 5:45
• A basic answer to your question (one that will fit in the comments), is that with a very large sample, you can detect an extremely small effect size. For example, if two groups differ on a 100 point test by 1 point (one group averages 90, the other 91), you probably won't be able to detect that with a sample of 20 people from each group. But you could detect it with samples of 200 from each group. I wouldn't say that the effect is necessarily less meaningful with large sample sizes, it's just that you may detect very, very small effects. – Duncan Apr 2 '15 at 16:34