Calculating Scale Scores with missing data
Locked
Calculating Scale Scores with missing data
 
|
Hi all,
I am in the process of cleaning my dissertation dataset. I was wondering what the command was to compute a sum score for participants who have 20% or less of the items missing. Currently I am working on the BDI and it has 21 items, my lab normally calculates scores for individuals who have 20% or less of the items missing on a measure. I thought I could do this with the following syntax (in the parenthesis are the 21 items in the BDI--however, I am trying to ask SPSS to calculate a score even in cases where the participants have up to 4 items missing). Can you tell me what I am doing wrong?
COMPUTE BDI=sum.17(BDI_1 + BDI_3 + BDI_4 + BDI_5 + BDI_6 + BDI_7 + BDI_8 + BDI_9 + BDI_10 + BDI_11 + BDI_12 + BDI_13 + BDI_14 + BDI_15 + BDI_16 + BDI_17 + BDI_18 + BDI_19 + BDI_19a + BDI_20 + BDI_21) .
EXECUTE. Many thanks! |
|
|
Hi Scott, For the SUM function you should have commas separating your variable
names, not plus signs. Hope this helps! Kylie. From: SPSSX(r) Discussion
[mailto:[hidden email]] On Behalf Of Scott Roesch Hi all, I am in the process of cleaning my dissertation
dataset. I was wondering what the command was to compute a sum score for
participants who have 20% or less of the items missing. Currently I am
working on the BDI and it has 21 items, my lab normally calculates scores
for individuals who have 20% or less of the items missing on a measure. I
thought I could do this with the following syntax (in the parenthesis are
the 21 items in the BDI--however, I am trying to ask SPSS to calculate a score
even in cases where the participants have up to 4 items missing). Can you
tell me what I am doing wrong? COMPUTE BDI=sum.17(BDI_1 + BDI_3 + BDI_4 + BDI_5 + BDI_6 +
BDI_7 + BDI_8 + BDI_9 + BDI_10 + BDI_11 + BDI_12 + BDI_13 + BDI_14 + BDI_15 +
BDI_16 + BDI_17 + BDI_18 + BDI_19 + BDI_19a + BDI_20 + BDI_21) . Many thanks!
|
|
|
First you need to contingency code beck19 (if they are trying to lose weight than any response should be set to 0, if they are not trying to lose weight than their score remains as is). Compute totalbeck=21*mean.19(bdi to bdi19, bdi20, bdi21). Note that bdi19a is not include. Martin sherman From: SPSSX(r) Discussion [mailto:[hidden email]] On Behalf Of Kylie Lange Hi Scott, For the SUM function you should have commas separating your variable names, not plus signs. Hope this helps! Kylie. From: SPSSX(r) Discussion [mailto:[hidden email]] On Behalf Of Scott Roesch Hi all, I am in the process of cleaning my dissertation dataset. I was wondering what the command was to compute a sum score for participants who have 20% or less of the items missing. Currently I am working on the BDI and it has 21 items, my lab normally calculates scores for individuals who have 20% or less of the items missing on a measure. I thought I could do this with the following syntax (in the parenthesis are the 21 items in the BDI--however, I am trying to ask SPSS to calculate a score even in cases where the participants have up to 4 items missing). Can you tell me what I am doing wrong? COMPUTE BDI=sum.17(BDI_1 + BDI_3 + BDI_4 + BDI_5 + BDI_6 + BDI_7 + BDI_8 + BDI_9 + BDI_10 + BDI_11 + BDI_12 + BDI_13 + BDI_14 + BDI_15 + BDI_16 + BDI_17 + BDI_18 + BDI_19 + BDI_19a + BDI_20 + BDI_21) . Many thanks! |
Re: Calculating Scale Scores with missing data
|
|
In reply to this post by Scott Roesch
Scott,
Just curious, is BDI referring to the Beck Depression inventory? ________________________________ From: SPSSX(r) Discussion on behalf of Scott Roesch Sent: Thu 1/20/2011 6:26 PM To: [hidden email] Subject: Calculating Scale Scores with missing data Hi all, I am in the process of cleaning my dissertation dataset. I was wondering what the command was to compute a sum score for participants who have 20% or less of the items missing. Currently I am working on the BDI and it has 21 items, my lab normally calculates scores for individuals who have 20% or less of the items missing on a measure. I thought I could do this with the following syntax (in the parenthesis are the 21 items in the BDI--however, I am trying to ask SPSS to calculate a score even in cases where the participants have up to 4 items missing). Can you tell me what I am doing wrong? COMPUTE BDI=sum.17(BDI_1 + BDI_3 + BDI_4 + BDI_5 + BDI_6 + BDI_7 + BDI_8 + BDI_9 + BDI_10 + BDI_11 + BDI_12 + BDI_13 + BDI_14 + BDI_15 + BDI_16 + BDI_17 + BDI_18 + BDI_19 + BDI_19a + BDI_20 + BDI_21) . EXECUTE. Many thanks! ===================== To manage your subscription to SPSSX-L, send a message to [hidden email] (not to SPSSX-L), with no body text except the command. To leave the list, send the command SIGNOFF SPSSX-L For a list of commands to manage subscriptions, send the command INFO REFCARD |
Re: Calculating Scale Scores with missing data
|
In reply to this post by Scott Roesch
Scott
Have a look at section 3.5 on page http://surveyresearch.weebly.com/block-3-analysing-two-variables-and-sometimes-three.html on my website. It's a fully worked set of syntax-based tutorials on the use of COUNT and COMPUTE with sets of attitude scales, together with commentary on whys and wherefores. Kylie is right about the commas, but if your items are contiguous in the file you could also write: compute bdi = sum.17 (bdi_1 bdi_3 to bdi_21) . Purists might argue that you should impute scores where they are missing: others might argue that you should only use scores from those who have a full set of responses, but this could well result in a very small sample! compute bdi = sum.21 (bd1_1 bdi_3 to bdi_21) . What exactly are the BDI items measuring? Are they on a 0 - n scale or a 1 - n scale? If the latter you can create a true zero point by subtracting the number of items in the scale: compute bdi = sum.21 (bd1_1 bdi_3 to bdi_21) - 21. John Hall [hidden email] http://surveyresearch.weebly.com ----- Original Message ----- From: Scott Roesch To: [hidden email] Sent: Friday, January 21, 2011 12:26 AM Subject: Calculating Scale Scores with missing data Hi all, I am in the process of cleaning my dissertation dataset. I was wondering what the command was to compute a sum score for participants who have 20% or less of the items missing. Currently I am working on the BDI and it has 21 items, my lab normally calculates scores for individuals who have 20% or less of the items missing on a measure. I thought I could do this with the following syntax (in the parenthesis are the 21 items in the BDI--however, I am trying to ask SPSS to calculate a score even in cases where the participants have up to 4 items missing). Can you tell me what I am doing wrong? COMPUTE BDI=sum.17(BDI_1 + BDI_3 + BDI_4 + BDI_5 + BDI_6 + BDI_7 + BDI_8 + BDI_9 + BDI_10 + BDI_11 + BDI_12 + BDI_13 + BDI_14 + BDI_15 + BDI_16 + BDI_17 + BDI_18 + BDI_19 + BDI_19a + BDI_20 + BDI_21) . EXECUTE. Many thanks! ===================== To manage your subscription to SPSSX-L, send a message to [hidden email] (not to SPSSX-L), with no body text except the command. To leave the list, send the command SIGNOFF SPSSX-L For a list of commands to manage subscriptions, send the command INFO REFCARD |
Re: Calculating Scale Scores with missing data (2)
|
|
In reply to this post by Scott Roesch
Scott
Left out the comma in my previous
reply
compute bdi = sum.17 (bdi_1 , bdi_3 to
bdi_21) .
I take it you're using the Beck Depression
Index?
You could also use cruder measures:
count crudebdi1 = bdi_1 , bdi_3 to bdi_21 (2,3) .
count crudebdi2 = bdi_1 , bdi_3 to bdi_21 (3) .
and see how they correlate with the more
precise bdi score
corr bdi crudebdi1 crudebdi2 .
or by
means bdi by crudebdi1 crudebdi2
.
|
Re: Calculating Scale Scores with missing data
|
In reply to this post by Scott Roesch
If for some reason you need the summative scores as a total something
like this untested syntax should work. It is the same as imputing the mean score for missing items. I would use this approach when I need the score as a total for some explicit comparative or other purpose. compute inflator = 21/nvalid(bdi1 to bdi21). compute bdi = mean.17(bdi1 to bdi21) * (inflator). However, a mean score and a total score have the same correlation with other variables. In some circumstances, rescaling to the response scale can be more intuitive. E.g., the mean of Likert items corresponds to the item response scale construct of strongly disagree to strongly agree. compute bdi = mean.17(bdi1 to bdi21). I also use the mean for extent scales (none or almost none to always or almost always), I use zero as the lowest value because of the cognitive correspondence of zero and none. To summarize (pun intended), to to keep the final summative scale score corresponding to the response scale, I tend to use the mean.n specification. without an inflator for missing items. Art Kendall Social Research Consultants On 1/20/2011 6:26 PM, Scott Roesch wrote: > Hi all, > I am in the process of cleaning my dissertation dataset. I was > wondering what the command was to compute a sum score for participants > who have 20% or less of the items missing. Currently I am working on > the BDI and it has 21 items, my lab normally calculates scores for > individuals who have 20% or less of the items missing on a measure. I > thought I could do this with the following syntax (in the parenthesis > are the 21 items in the BDI--however, I am trying to ask SPSS to > calculate a score even in cases where the participants have up to 4 > items missing). Can you tell me what I am doing wrong? > COMPUTE BDI=sum.17(BDI_1 + BDI_3 + BDI_4 + BDI_5 + BDI_6 + BDI_7 + > BDI_8 + BDI_9 + BDI_10 + BDI_11 + BDI_12 + BDI_13 + BDI_14 + BDI_15 + > BDI_16 + BDI_17 + BDI_18 + BDI_19 + BDI_19a + BDI_20 + BDI_21) . > EXECUTE. > Many thanks! > > ===================== To manage your subscription to SPSSX-L, send a message to [hidden email] (not to SPSSX-L), with no body text except the command. To leave the list, send the command SIGNOFF SPSSX-L For a list of commands to manage subscriptions, send the command INFO REFCARD
Art Kendall
Social Research Consultants |
|
|
It's been a while since I worked with BDI data and that was
BDI ver 1 (I believe that there is a BDI ver 2 in use now, so it probably is helpful to identify which version is being used; offhand I don't know how ver 2 differs from ver 1). A few more points: (1) The original BDI was not a Likert scale and for clinical purposes the sum of the items is compared to certain "thresholds" to identify whether a person has a "clinically significant" degree of depression (there is a correlation between BDI sum and whether one would receive a DSM diagnosis of Major depression or other depressive diagnosis). One might be interested in the mean of the scale for other reasons but, unless the ver 2 manual suggests using the mean of all BDI scores for a particular purpose, I think most people would not be interested in it. (2) It would probably be useful to determine the descriptive statistics for the number of missing items (e.g., minimum number, maximum number, mode(s), etc.), whether certain items appear to be consistently missing (i.e., what is their content, why would the items not be responsed to, etc.), and is there a pattern of correlation among the missing items, that is, are the items missing "apparently" at random or are certain items consistently not responded to. The next set of questions would be concerned with who provides complete responses, who misses a few responses, and who does not respond to a large number of items (according to your criterion, >20% missing would be considered "large"). Factors such as familiarity with the BDI (e.g., novelty) may affect responses as well as possibly severity of illness and/or number of lifetime episodes of depression and/or other illnesses. -Mike Palij New York University [hidden email] ----- Original Message ----- From: "Art Kendall" <[hidden email]> To: <[hidden email]> Sent: Friday, January 21, 2011 8:05 AM Subject: Re: Calculating Scale Scores with missing data > If for some reason you need the summative scores as a total something > like this untested syntax should work. > It is the same as imputing the mean score for missing items. > I would use this approach when I need the score as a total for some > explicit comparative or other purpose. > > compute inflator = 21/nvalid(bdi1 to bdi21). > compute bdi = mean.17(bdi1 to bdi21) * (inflator). > > However, a mean score and a total score have the same correlation with > other variables. > In some circumstances, rescaling to the response scale can be more > intuitive. E.g., the mean of Likert items corresponds to the item > response scale construct of strongly disagree to strongly agree. > compute bdi = mean.17(bdi1 to bdi21). > I also use the mean for extent scales (none or almost none to always or > almost always), I use zero as the lowest value because of the cognitive > correspondence of zero and none. > > To summarize (pun intended), to to keep the final summative scale score > corresponding to the response scale, I tend to use the mean.n > specification. without an inflator for missing items. > > Art Kendall > Social Research Consultants > > > On 1/20/2011 6:26 PM, Scott Roesch wrote: >> Hi all, >> I am in the process of cleaning my dissertation dataset. I was >> wondering what the command was to compute a sum score for participants >> who have 20% or less of the items missing. Currently I am working on >> the BDI and it has 21 items, my lab normally calculates scores for >> individuals who have 20% or less of the items missing on a measure. I >> thought I could do this with the following syntax (in the parenthesis >> are the 21 items in the BDI--however, I am trying to ask SPSS to >> calculate a score even in cases where the participants have up to 4 >> items missing). Can you tell me what I am doing wrong? >> COMPUTE BDI=sum.17(BDI_1 + BDI_3 + BDI_4 + BDI_5 + BDI_6 + BDI_7 + >> BDI_8 + BDI_9 + BDI_10 + BDI_11 + BDI_12 + BDI_13 + BDI_14 + BDI_15 + >> BDI_16 + BDI_17 + BDI_18 + BDI_19 + BDI_19a + BDI_20 + BDI_21) . >> EXECUTE. >> Many thanks! >> >> > > ===================== > To manage your subscription to SPSSX-L, send a message to > [hidden email] (not to SPSSX-L), with no body text except the > command. To leave the list, send the command > SIGNOFF SPSSX-L > For a list of commands to manage subscriptions, send the command > INFO REFCARD ===================== To manage your subscription to SPSSX-L, send a message to [hidden email] (not to SPSSX-L), with no body text except the command. To leave the list, send the command SIGNOFF SPSSX-L For a list of commands to manage subscriptions, send the command INFO REFCARD |
Re: Calculating Scale Scores with missing data
|
|
Thanks for the clarification. Published cut-offs, percentile transform
tables, previous studies are the kind of comparative purposes I should have elaborated. Art Kendall Social Research Consultants On 1/21/2011 8:40 AM, Mike Palij wrote: > It's been a while since I worked with BDI data and that was > BDI ver 1 (I believe that there is a BDI ver 2 in use now, so > it probably is helpful to identify which version is being used; > offhand I don't know how ver 2 differs from ver 1). > > A few more points: > > (1) The original BDI was not a Likert scale and for clinical > purposes the sum of the items is compared to certain "thresholds" > to identify whether a person has a "clinically significant" degree > of depression (there is a correlation between BDI sum and > whether one would receive a DSM diagnosis of Major depression > or other depressive diagnosis). One might be interested in the > mean of the scale for other reasons but, unless the ver 2 manual > suggests using the mean of all BDI scores for a particular purpose, > I think most people would not be interested in it. > > (2) It would probably be useful to determine the descriptive > statistics for the number of missing items (e.g., minimum number, > maximum number, mode(s), etc.), whether certain items appear > to be consistently missing (i.e., what is their content, why would > the items not be responsed to, etc.), and is there a pattern of > correlation among the missing items, that is, are the items missing > "apparently" at random or are certain items consistently not responded > to. The next set of questions would be concerned with who provides > complete responses, who misses a few responses, and who does > not respond to a large number of items (according to your criterion, >> 20% missing would be considered "large"). Factors such as > familiarity with the BDI (e.g., novelty) may affect responses as > well as possibly severity of illness and/or number of lifetime episodes > of depression and/or other illnesses. > > -Mike Palij > New York University > [hidden email] > > > > ----- Original Message ----- > From: "Art Kendall"<[hidden email]> > To:<[hidden email]> > Sent: Friday, January 21, 2011 8:05 AM > Subject: Re: Calculating Scale Scores with missing data > > >> If for some reason you need the summative scores as a total something >> like this untested syntax should work. >> It is the same as imputing the mean score for missing items. >> I would use this approach when I need the score as a total for some >> explicit comparative or other purpose. >> >> compute inflator = 21/nvalid(bdi1 to bdi21). >> compute bdi = mean.17(bdi1 to bdi21) * (inflator). >> >> However, a mean score and a total score have the same correlation with >> other variables. >> In some circumstances, rescaling to the response scale can be more >> intuitive. E.g., the mean of Likert items corresponds to the item >> response scale construct of strongly disagree to strongly agree. >> compute bdi = mean.17(bdi1 to bdi21). >> I also use the mean for extent scales (none or almost none to always or >> almost always), I use zero as the lowest value because of the cognitive >> correspondence of zero and none. >> >> To summarize (pun intended), to to keep the final summative scale score >> corresponding to the response scale, I tend to use the mean.n >> specification. without an inflator for missing items. >> >> Art Kendall >> Social Research Consultants >> >> >> On 1/20/2011 6:26 PM, Scott Roesch wrote: >>> Hi all, >>> I am in the process of cleaning my dissertation dataset. I was >>> wondering what the command was to compute a sum score for participants >>> who have 20% or less of the items missing. Currently I am working on >>> the BDI and it has 21 items, my lab normally calculates scores for >>> individuals who have 20% or less of the items missing on a measure. I >>> thought I could do this with the following syntax (in the parenthesis >>> are the 21 items in the BDI--however, I am trying to ask SPSS to >>> calculate a score even in cases where the participants have up to 4 >>> items missing). Can you tell me what I am doing wrong? >>> COMPUTE BDI=sum.17(BDI_1 + BDI_3 + BDI_4 + BDI_5 + BDI_6 + BDI_7 + >>> BDI_8 + BDI_9 + BDI_10 + BDI_11 + BDI_12 + BDI_13 + BDI_14 + BDI_15 + >>> BDI_16 + BDI_17 + BDI_18 + BDI_19 + BDI_19a + BDI_20 + BDI_21) . >>> EXECUTE. >>> Many thanks! >>> >>> >> ===================== >> To manage your subscription to SPSSX-L, send a message to >> [hidden email] (not to SPSSX-L), with no body text except the >> command. To leave the list, send the command >> SIGNOFF SPSSX-L >> For a list of commands to manage subscriptions, send the command >> INFO REFCARD ===================== To manage your subscription to SPSSX-L, send a message to [hidden email] (not to SPSSX-L), with no body text except the command. To leave the list, send the command SIGNOFF SPSSX-L For a list of commands to manage subscriptions, send the command INFO REFCARD
Art Kendall
Social Research Consultants |
«
Return to SPSSX Discussion
|
1 view|%1 views
| Free forum by Nabble | Edit this page |