z-score used as X axis variable

#1
Seeking your opinion on the use of z-scores instead of original data for a correlation graph.
I have a paper showing a graph of two factors, that should correlate because the authors state that the multiple regression result shows that only variable X data explains changes in Y data.
However, instead of plotting the actual X data the authors have chosen to plot the z-scores. Would z-scores exaggerate the change in X values so the graph of the relationship with Y appears more convincing? Even on this graph it looks like the regression line is being riven by three extreme points, while the remaining 20 points appear completely random.
Looking forward to your advice.
Nayo
 
#2
This is what I want to see and would be grateful for confirmation if this is a logical request.
The graph of variable X vs variable Y with the regression line calculated from the least squares of the residuals.
Please let me know what you think,
Nayo
 

hlsmith

Not a robit
#4
Was the variable standardized in the multiple regression model - a standard deviation change in X associated with a ?? change in Y?

Also, I agree with @Buckeye - provide link to "paper".
 
#5
Hi Buckeye and hlsmith,
Thank you for your supportive replies. Sadly I am not permitted to post their graph at this time. I have attached a very similar graph. The Y axis has straightforward raw data plotted. The X axis has the z-scores of the other variable plotted on a scale from -2 standard deviations through zero to +2 SD. To answer your question, yes, the authors state this variable is standardised. Ten data points fall within the confidence intervals lines for the regression line, thirteen outside. The image I have attached is very similar in the spread of data.
The results section consists of two sentences. The first states multiple regression was used to evaluate the association between X and Y. The second gives the results showing a statistically significant positive relationship between X and Y values and that other factors did not contribute to the equation. I have no problem with the use of scatter plots with the regression line and confidence intervals. What I query is plotting the z-scores without explanation in the text or in the legend that accompanies the graph.
I really appreciate your time in considering this. I look forward to your advice.
Nayo
 

Attachments

hlsmith

Not a robit
#7
Are you peer reviewing this paper? Yeah it could be confusing changing scale and a better description and labeling is probably needed. I can't remember the term, but I believe standardizing data is a "monotonic" transformation meaning it does change result when viewing conclusions. I would like to confirm that, I can't think why the results would differ! Though remember I am not providing any citations!!!