correlation diagram with motor severity score Y using a graph builder by adjusting biomarker A according to age and gender

[#1]

We would like to clarify whether there is a linear relationship between biomarker X(DATQUANT) and motor severity score Y (MDS-UPDRS 3 total score) in patients with a neurological disorder. In a simple correlation, biomarkerX  is associated with motor severity score Y.Since age and biological sex influence the biomarker X, we considered them as covariates. 

At that time, I am thinking of creating a correlation diagram with motor severity score Y using a graph builder by adjusting biomarker A according to age and gender.

 We would appreciate it very much if you could let us know how to obtain scatter plot of age- and gender-adjusted biomarker X vs motor severity score Y using JMP.  

[#2]

I just took a quick look at the data provided and some of your questions.  Here is what I see in the data and some questions:

1. I don't have any context for the data.  How much of a change in DAT or MDS is of practical significance?

2. Regarding correlation between MDS and DAT, there is not a strong correlation (r=-0.32) and there are a number of potential outliers (Mahalanobis, 7).  This is not accounting for the age/sex effects.

3. Looking at the relationships between DAT and age and sex, it appears sex has almost no relationship (fit model).  Age is statistically significant.  Those two terms account for only about ~20% of the variation in the data.  As age increases, DAT decreases.

4. There does appear to be a linear relationship between DAT and MDS, but in that model both Age and Sex are insignificant.  The model containing all three terms accounts for ~8% of the variation in the data.

How confident are you in the measurement systems?

[#3]

Hi @KF2 ,

One possibility is to regress biomarker X on age and gender, save the residuals and then plot them against motor severity score Y in graph builder as a simple scatterplot. This is what that looks like with the data you provided,

Another possibility, for treating age and sex as covariates, is to do a hierarchical regression. In other words, you can fit a regression model where age and sex predict your outcome, then fit another regression model where age, sex, and biomarker X predict your outcome. You'll obtain the R-squared for each of those models (turns out it's .002 and .11, respectively). You can then look at the "Sequential Tests" (or Type I Sum of Squares) of the effects to obtain a test for the change in explained variance from one model to the next. In this case, the change is statistically significant with an F-value of 11.23, p = .001.

I saved the scripts of each of these steps in the attached data table.

HTH,

~Laura

[#4]

Thank you for your reply.

We would like to clarify whether there is a linear relationship between biomarker X (DATQUANT: dependent variable) and motor severity score Y (MDS-UPDRS-3: explanatory variable). Since age and biological sex influence the biomarker X, we considered them as covariates. We would like to conduct multiple regression.

[X (DATQUANT: dependent variable)] = a1*Y (MDS-UPDRS-3: explanatory variable) + a2*[age] + a3 + residuals

a1, a2 : slope coefficients for each explanatory variable

a3: constant term for biological sex

Now we would like to visualize the relationship between [age and sex adjusted (DATQUANT: dependent variable)] and Y (MDS-UPDRS-3: explanatory variable)

So we would like to prepare scatter plot

adjusted X (a1*[X(DATQUANT: continusous variable)] - a2*[age] - a3) vs. Y (MDS-UPDRS-3: dependent variable)

We would appreciate it very much if you could give us comments and suggestions.

Many thanks in advance.