In brief overview, I have grown some plants and measured plant height (amongst many other traits), but my plants have been damaged by frost. I gave each plant a frost score between 0-5 as a measure of the damage. Therefore, I'd like to know how I can adjust my plant height data in such a way to remove the effect of frost, using my frost score.
It was suggested to me that I might be able to run a regression analysis and, using the linear equation, determine "as frost score increases by X, plant height decreases by Y" and thereby create a correction formula to apply to plant height data. Can anyone recommend a method to doing this using SPSS?
To clarify, for example: if plant height decreases by 0.5 cm as frost score increases by 1, then a recorded height of 10 cm will be corrected to 9 cm if the frost score is 2.
Hello Jemma, I think you can regress Plant Height (dependent variabile) against Frost score (independent variable) and save the Unstandardized Predected Values to be used in subsequent analyses (instead of raw plant height).
In SPSS vers. 23 go to Analyze => Regression => Linear then choose "Plant Height" as Dependent, Frost score ad Independent; press the "Save" button option on the right and choose "Predicted Values" Unstandardized. Run the analysis (click OK).
If you prefer use the syntax:
REGRESSION
/MISSING LISTWISE
/STATISTICS COEFF OUTS R ANOVA
/CRITERIA=PIN(.05) POUT(.10)
/NOORIGIN
/DEPENDENT PH
/METHOD=ENTER FS
/SAVE PRED.
Where "PH" is Plant Height and "FS" is Frost score.
Kind regards
Vincenzo
The frost score should be negative since it decreases plant heights. Also you need a few undamaged plants with frost score zero. Best, David
@Vincent Calvo
Hello and thank you for your suggestion.
As per your guidance, I ran the linear regression such that it would give me the Unstandardised Predicted Values, but upon looking at the resulting column with the new values, I'm not sure this has worked the way I intended - for example, if the plant height value was 10 cm and the frost score was 0, I would have expected the new value to also be 10 cm (as there has been no damage and therefore the height value needn't be adjusted) but this was not the case (e.g. frost score of 0 resulted in 10 cm being changed to either just above or just below 10 cm). As such, I'm not sure the Unstandardised Predicted Values are useful for further analyses in my case, but thank you nonetheless, I was previously not aware of the functions you have mentioned and am glad to know they are there.
@David Eugene Booth
Thank you for the suggestion. I ran the linear regression as per Vincent Calvo's post, but instead using negative frost scores (as compared to positive in the first regression I ran) and the values have come out identical, thus using negative frost scores doesn't seem to affect the adjusted plant height values. Furthermore, I do indeed have some undamaged plants that have a frost score of 0, but as I mentioned in my reply to Vincent Calvo, obtaining Unstandardised Predicted Values from the linear regression does not seem to be the way to go, as plant height values were being adjusted even if frost score was 0.
If you have any further recommendations, I would greatly appreciate it.
- Badges
- Science topic
- Similar topics
- Mathematics
- Statistics
- Regression