Hello I have completed a study examining how training effects employees intention to remain in the organisation. Below are the relevant variables and information needed:
Question
1. What is the best SPSS tool to run analysis ?
2. The hypothesis 1 and 2 are comparing conditions (significance and difference) do i have to reconstruct my hypothesis or is there a statistical tool i can use?
Independent variable:
1. Training specificity - (3 conditions: Specific training, General training and combination training) Nominal data
2. Training location (3 conditions: Off-the-job training, on-the-job training and combination location) Nominal data
Dependent variable:
Intention to remain in the organisation - Ordinal data
Meditators
- Affective organisational commitment:
- Normative organisational commitment
- Perceived Psychological contract breach.
Control variables: Age, Gender, Education level, employment status, employment status, frequency and organisation tenure.
These are my hypothesis:
H1 -
Specific training and Combination training will be related to intention to remain, whereas general training will be associated with lower levels of intention to remain.
H2
On-the-job training and combination training will be positively related to intention to remain, in comparison to off-the-job training
H3
Affective & Normative commitment will mediate the relationship between training and intention to remain.
H4
Perceived psychological contract breach will mediate the relationship between training and intention to remain within the NPO.
Thank you so much for taking the time to read this. This is for my dissertation and I am really confused!
Hi Gayatri Dogra,
First, let me re-state what you wrote: You basically have a 3 (Training Specificity: Specific, general, combination ) x 3 (Training location: Off-the-job , on-the-job, combination) design. Essentially, two of the conditions are a mixture of the other ones (the two combination conditions). This does pose some problems, since they are obviously more powerful conditions because they combine the quality of the other two. But this should not limit your ability to interpret them, even though I am not really sure why you included them (managerially, they might have interesting implications, even though I would expect this also to be an issue of budget available). Your dependent variable is the intention to remain in the organisation. I am assuming that you measured it with a Likert-type scale (5 point, 7 point, 9 point, etc.?). This would mean that you can at least treat it as approximately interval-scaled (even though this is - in terms of measurement theory - not really the case). Your mediators are probably also measured on a Likert-type scale (you did not provide information on that).
Second, the typical way to go about the analysis would be with conditional process modeling (Andrew Hayes is one of the main people promoting this approach, http://afhayes.com/introduction-to-mediation-moderation-and-conditional-process-analysis.html). There is also a book and a myriad of articles about this approach. It is basically a regression-based approach with bootstrap inferences. This allows you to probe the confidence intervals of your effects. Here, you can include multiple moderators, you can include mediators, etc. I would probably analyze the models one mediator at a time, before you decided to potentially include multiple mediators at once (this is also possible). Version 3 of his PROCESS macro allows for the use of independent variables and moderators with more than two categories (as in your case).
Third, regarding your hypotheses, I would not phrase them in a way that uses "relate". Rather, you might think about contrasting the expected effects. For example, H2 could instead be formulated such that "Intention to remain with the company is higher for participants in the "On-the-job training" and "combination training" condition as compared to the "off-the-job training" condition. Similarly, with the mediating hypotheses, I am also missing a sense of direction. For example (and I am just making the directions up here), H3 could be formulated such that you first introduce the expected impact of the independent variables on the mediators and then introduce the expected mediation. Otherwise, you are missing one important part of the proposed causal link between the three variables (IV, Mediator, DV). Also, the question would be if there is also a moderated mediation, such that Training Location and Training Specificity interactively impact the mediator or the DV through the mediator. This will get quite complex, but I suppose you have some theory to derive your expectations.
I hope this help, please let me know.
Best,
Alexander
Alexander Mafael Thank you so much for the taking out time to respond to my question. Firstly to address the questions/points raised
1. I wanted to explore combination (in training specificity variable, combination essentially is that participants attended both general and specific training); empirically this has not been explicitly examined in the past.
2. Yes my dependent variable and meditators are measured on a 5-point Likert Scale.
3. The conditional process modeling is a great suggestion. I explored this todya and tested the interaction between IV and DV including control variables, and IV and DV + control variables + meditators (separately). If you are interested in seeing the results I can post that too, to see if I am conducting the test properly.
4. I really like how you phrased the hypothesis, thank you for that.
5. When formulating the hypothesis for meditators (the literature doesn't explicitly demonstrate the direction just significance), do you suggest something like this, e.g. H4 - Perceived psychological contract breach will negatively meditate the relationship between training specificity and intention to remain ?
Follow-up questions
1. When constructing the hypothesis for meditators is it important that i keep training conditions separate(specificity and location) or is it ok to combine ?
2. When I ran the test for conditional process modeling, it only shows the significance of the variables training specificity/ training location rather than the categories(e.g. general, specific, combination). As a result I am confused about how I can justify the significant differences between the categories to show which category has the highest and lowest effect on the DV. I attempted to initially answer this question by thinking, perhaps the means and/or standard deviation could indicate; but then another question arises as to how to distinguish, how significantly different the categories are.
I look forward to hearing your response.
Best wishes
Gayatri
* Bold ~ very important question
Hi Gayatri Dogra ,
Regarding the mediation hypotheses, this really depends on whether you expect the mediator to have different effects at levels of the independent variables. That is, do you think that (for example) affective organizational commitment will mediate the effect of training specificity for general, but not for specific training? This would imply a moderated mediation. In a moderated mediation, the mediating effect differs across levels of the moderators. If you do not have any a priori hypotheses regarding this, it might also be OK to take a more explorative stance on the reporting of different effects.
Regarding your bold question: Similar to an ANOVA, the significance simply indicates that there is an overall effect of the IV on the DV. To investigate the effects in the different conditions you would need to look at the rest of the output, which usually specifies whether the effect is different for other conditions of the IV. Note that the main effect than represents the effect of your baseline condition (depending on how you coded your data). Here is a simple, but illustrative video on moderator analysis in PROCESS: https://www.youtube.com/watch?v=Zx1DJfhzO_o
Hope this helps!
Best,
Alexander
- Badges
- Science topic
- Similar topics
- Mathematical Sciences
- Statistics
- Statistical Software
- SPSS