Analysis of Data Part 1

Of variables and measurements

We're interested in conducting our own investigation to see if arm-span is indeed correlated with height, and if it does, what is the nature, significance and magnitude of this association.

Based on our literature review, we saw there is a clinical relevance of estimating height of a person using the arm-span. This estimation is useful in the following cases -

• to quantify age-related loss in stature


• to identify clients with disproportionate growth abnormalities and skeletal dysplasias


• to quantify the alteration in height as a result of progressive spinal deformities


• as a substitute measurement where height measurement is not feasible, such as clients' lower limbs have been amputated

Therefore, for the purposes of statistical analysis, we've identified arm-span to be our independent variable, and height, the dependent variable.

The following data elements are collected in the course of measuring our sampling units.

1. age (level of measure: scale)


2. gender (level of measure: nominal, where 1='female', 2='male')


3. height (level of measure: scale)


4. armspan (level of measure: scale)

Sampling method

We've not employed probability sampling in our selection of sampling units. Our chosen sampling method is coincidental sampling. We've allocated for our group one week to collect the necesary data, and on day 1, we went around the class during 2 15-minute breaks in between lessons, asking for classmates' permission to allow our group to measure their height and arm-span.



Three timeslots are identified over the next 4 days from which they are welcome to indicate their preference to turn up for the measurement process. In this manner, we managed to enlist the help of 26 classmates, and including the 4 of us, we had the minimum sample size of 30.




In a subsequent literature review, we found the rationale for why Ms Chia has stipulated 30 as the minimum sample size. Creswell (2008, p.370) explained that "The group needs to be of adequate size for use of the correlational statistic, such as N = 30; larger sizes contribute to less error variance and better claims of representativeness."


We looked up on this internet site recommended by Ms Chia, during one of the research methods' lecture. (http://www.custominsight.com/articles/random-sample-calculator.asp) Using the calculator provided, and given a population size of 46 that is the total size of our cohort for HSNF01, our sample size of 30 yields a sampling error of 8.9%, 10.6%, and 13.9% at 90%, 95% and 99% confidence interval respectively.