Indian Journal of Science and Technology
DOI: 10.17485/ijst/2016/v9iS1/104115
Year: 2016, Volume: 9, Issue: Special Issue 1, Pages: 1-6
Original Article
Anandakuttan B. Unnithan*
IIM Kozhikode, Kozhikode - 673570, Kerala, India; [email protected]
*Author for correspondence
Anandakuttan B. Unnithan
IIM Kozhikode
Email:[email protected]
Objectives: This study compares the scale characteristics when the same instrument is used with five and four point scales to test the assumption whether results vary with number of scale points. Methods/Statistical Analysis: An experimental study was designed where the same items were organized into two different forms one with five point response format and the second with four points format. 195 respondents were randomly assigned to either one of the formats. Comparison was made between the groups on the scale characteristics and correlations obtained. Findings: When the data is transformed to a common scale, it is seen, that data characteristics like mean, standard deviation, skewness and kurtosis are comparable indicating that there is no change in scale characteristics. However when the correlations are estimated, it is seen that five point scales report a significantly higher correlation. Application/Improvements: The major implication is that five point scales tend to inflate the reported correlations with a distinct chance of higher type1 error especially when the variables may not be associated.
Keywords: Correlation, Five Point Scale, Likert Scale, Measurement, Scale Points, Type 1 Error
Subscribe now for latest articles and news.