Dear Allstat =
1. I would much welcome advice on how to conduct the Split-Half Reliability test on mail questionnaire survey data using SPSS v 14, assuming it is possible to do so. Others well-known options such as Test/Retest are not feasible.
2. I have 70 cases and some 130 variables (almost entirely IVs) of mixed types: dichotomous (yes/no) 39 per cent; other categorical (including 3/4 options and open-ended) 43 per cent; numerical/ continuous 15 per cent; and Likert scales 3 per cent.
One scale has 5 intervals; the others 11 each. All are single-item scales that show one score per case. Variables of the different types were well mixed throughout the questionnaire ( possibly helping to offset common method variance) and this variety is reflected in the SPSS files. Almost all variables comprise factual data relating to company trading performance.
3. None of the practical examples I have yet seen in Net articles or elsewhere appear to deal with mixed data such as mine. Those for both Split-Half and Cronbach Alpha seem to to focus solely on multi-item scales.There is much discussion about the merits and demerits of different kinds of split. There is also a recognition that dichotomous variables need specialised treatment. But nothing that I have seen casts light on how I should proceed to test the reliability of my particular dataset.
I shall be obliged for any advice or assistance that listers can offer towards the resolution of the above problem. With advance thanks =
Owen
|