# Is there any sequence dissimilarity measure that has an intuitive interpretation like the dissimilarity index has?

I want to know how much the sequences in my sample differ from a given ideal-typical sequence. Is there any intuitive way of interpreting the dissimilarity measures for sequences?

If it would be the dissimilarity index from the segregation literature, for example, I would be able to interpret an index of 0.45 as 45% of the people of one racial group having to move in order to achieve a perfectly even distribution over the neighbourhoods. Are there ways to interpret dissimilarities between sequences in a similarly intuitive way?

Here is a bit of background for asking this question. I have a particular type of sequence I want to study on an immigrant sample. The idea is to see how much the sequences of immigrants differ from the sequences of natives and what are the covariates that can explain this difference. Having an interpretable dissimilarity measure becomes important when describing the associations between covariates and the dissimilarities (e.g. 'a unit increase in X makes your sequence k states more similar to the reference sequence' or 'a unit increase in X makes your sequence k% more similar to the reference sequence').

Any suggestions on how to do this?

• From what I understand, the dissimilarity index you are referring to measures the dissimilarity between two groups (or regions), while the dissimilarity between two sequences compares individuals, not groups. Anyway, each dissimilarity measure has its own intuitive interpretation. Also, we should be able to use any dissimilarity measure to evaluate the impact of an increase in a covariate X on the similarity with a reference sequence. Why would that depend on the measure chosen? Feb 16, 2018 at 14:51
• If I’d be using Duncan’s D, a change in x% would mean that x% people would have to shift groups to obtain an even distribution. What would the interpretation of an effect that changed the dissimilarity between sequences by x? Is there any way to interpret them as something like “the effect is X size, which means that the Y states in the sequences would become similar,” or the like. Something that can be translated into real world units. Feb 16, 2018 at 14:56