# Handling non-significant beta coefficients

In our thesis, we have asked participants (consumers) about how several independent factors/variables (like price) affect the willingness to buy (the dependent factor). We have used the beta coefficient to analyse the data. However, some beta coefficients are not significant. Any tips on what we can do with the data to change this? Alternatively, how can "non-significant data" be presented in the thesis? Furthermore, can we use the R-squared score even though some beta coefficients used in its calculation were not significant?

Thanks for any help. :)

Any tips on what we can do with the data to change this?

Why change this? A null result is still a result, and if you purposefully seek to find data to support your hypothesis, that isn't really how science is done. My answer here may also be relevant.

• (1) Good answer: Succinct and to the point. Commented Feb 5 at 21:11

This often happens, moreso with observational data.

If variables were included to test specific hypotheses, then you have your answer.

If variables were included because past research showed they were important, then you probably still have an answer. At least in the circumstances you are investigating these variables are not significant.

There are several issues here. One is the source and nature of your data, which could be different from previous research. Another possibility is that you have used different variables from prior research. You might have included more variables and some of them account for any effect. Your variables might have been measured differently, eg, different scales, and have different effects.

You might just have a small sample size relative to other research or more variation in your dependent variable than others.

Any of these possibilities should be part of your written research plan and your discussion of results. You had a reason to include the nonsignificant variables and seem somewhat surprised by this. This deserves explanation.

If you think there is a good reason to examine your model after removing the insignificant variables you can go ahead. Plan to examine and understand what has happened.

You might have correlated independent variables, either in pairs or in linear combinations, ie, multicollinearity. You should at least examing the correlation matrix of the independent variables to see if there are any obvious relationships that you did not anticipate.

If you have included sets of variables that measure essentially the same thing you might want to take some of them out selectively since these are likely to be multicollinear.

Likely you should report the model you already have and any model that results from removing some of your independent variables. This is not only honest but could be informative about the overall relationshipl Any focus on R-squared is misplaced. This is a useful descriptive statistic of the success of the overall regression and not much else. Your R-squared will decrease when you remove some variables but not by much. The R-squared and adjusted R-squared will be closer together.