How many levels are there in your variable and how are they defined?

Great question! Since your predictor variable (true crime consumption) is measured on a 5-point Likert scale, using sequential coding is a valid approach to preserve the ordinal information without assuming equal intervals as in continuous data. Each coded variable compares adjacent levels, and the coefficients in your moderation model show how moving from one level to the next affects sleep quality, depending on sensitivity to violent content. In your write-up, you might say:

“Sequential coding was used to model ordinal changes in true crime consumption, allowing for the interpretation of incremental effects within the moderation framework.”

Thank you for your reply! There are 5 levels and they are defined as never, once a month, 2-3 times a month, 1-3 times a week, or every night.

Amaal Mohammad Khalil Alshaar thank you for your help!! Will the use of coding change the way I interpret the data/report the moderation effect? As my SPSS output now has interaction effects for the different levels rather than just one interaction effect.

When you use sequential coding for an ordinal variable with 5 levels, SPSS treats each adjacent level as a separate comparison. This results in multiple interaction terms, each revealing how the moderation effect differs between specific levels—instead of a single overall effect.

I was curious—what led you to choose sequential coding over treating the variable as continuous? Was this decision based on theoretical reasoning, supervisor guidance, or the specific goals of your research?

Amaal Mohammad Khalil Alshaar Thank you for helping! Will I have to report each individual moderating effect in my results section?

I chose to do it as I had a conversation with my supervisor, and she said that my variable was categorical, so it wouldn't be able to be used in the analysis as it is, and my groups do not have equal values between them. Is this the correct thing to do?

Izzy Waller Because you gave intervals for each level you could recode each of those to its midpoint, and then treat that as an interval variable measured in "times per month."

It may be easier to help you if you let us know what model you are using . In spss process models there is provision for catagorical models for moderated relationships . Check models 8 and 59 in Haynes process models .

David L Morgan Thank you! will this cause an issue as there is not equal intervals between each of the categories?

No, there is ne need for equal intervals. As an example, consider converting different levels of education into years of schooling.

Sequential coding (also known as Helmert coding) is a method of contrast coding often used when your predictor variable is ordinal or multicategorical (like your 5-point Likert scale for true crime consumption). Rather than assuming equal spacing (as in treating the variable as continuous), it creates contrasts between each level and the mean of the previous levels.

For example, with a 5-point scale (e.g., 1 = Never to 5 = Very often), the sequential coding will create 4 new variables:

• Contrast 1 compares Level 2 vs Level 1
• Contrast 2 compares Level 3 vs mean of Levels 1 and 2
• Contrast 3 compares Level 4 vs mean of Levels 1–3
• Contrast 4 compares Level 5 vs mean of Levels 1–4

This approach respects the ordering of the categories without assuming equal intervals, making it suitable for ordinal predictors in moderation analysis.

In your moderation model (e.g., using regression), each interaction term (e.g., Contrast × Moderator) tells you whether the moderator (sensitivity to violent content) changes the effect of a specific contrast in true crime consumption on sleep quality.

For instance:

• A significant interaction on Contrast 3 means the difference in sleep quality between Level 4 and the average of Levels 1–3 depends on the level of sensitivity to violent content.

So instead of interpreting a single slope (as you would with a continuous predictor), you interpret each contrast separately. This can help uncover more nuanced patterns—for example, maybe the relationship between true crime and sleep is only affected by the jump from “Sometimes” to “Often.”

In your results section, you could write something like:

To account for the ordinal nature of the predictor variable (true crime consumption), sequential (Helmert) coding was used. Four contrasts were created, each comparing a level of consumption to the mean of the preceding levels. Each contrast and its interaction with sensitivity to violent content were entered into the moderation model. A significant interaction on Contrast 3 indicated that the moderating effect was particularly evident when comparing frequent consumption to lower levels (Levels 1–3), B = ..., p = .... Simple slopes analyses were conducted to probe these interactions.

Optionally, include a plot of the interactions to make it more digestible.

Hi! Great to hear about your research. It’s a very relevant and interesting topic.

You're absolutely right that moderation analysis traditionally assumes continuous predictors, but using ordinal predictors (like a 5-point Likert scale) is common in psychology, especially when properly coded. Since you've applied sequential (or dummy) coding, you're essentially treating the Likert scale as a categorical variable.

In terms of reporting:

• State clearly in your methods section that the predictor was ordinal and that sequential coding was applied to create contrast variables.
• In your results section, report the coefficients for each coded variable (e.g., comparing level 2 to level 1, level 3 to level 2, etc.), and interpret each contrast in terms of how the relationship between true crime and sleep quality changes across those consumption levels.
• If sensitivity to violent content is your moderator, describe whether and how the strength or direction of the relationship differs at each coded level.

Also, justify your choice by citing that ordinal Likert-type data, especially with 5+ points, can be analyzed using regression-based techniques when assumptions are handled carefully (see Norman, 2010 for a good source on this).