question on ttest

[Deepa Lakshmi Shanadi]

Dear all,

 

1) I would like to do a ttest trying to see if there is a difference between headteachers and teachers  in their means on over/underreporting attendance.

 

2) Each HT and teacher belong to a intervention and control group.

 

3) HT and Teacher variables are continuous variables. There are negative numbers in the discrepancy colum (negative- underreporting, positive number- overreporting).

The syntax I have so far is:

 

T-TEST GROUPS=SCHOOLFEEDING(1 2)
  /MISSING=ANALYSIS
  /VARIABLES=Discrepancyteacher1 DiscrepancyHT1
  /CRITERIA=CI(.95).

However when I get my output, it looks like it is tell ing me if there is a difference between teachers in the intervention and control group and a difference between headteachers in the intervention and control group.

 

Am I using the wrong type of ttest?  Is this the ttest that I need- see below? And what do I need to be careful of with the negative numbers in the column? Thanks so much as usual?

 

USE ALL.
COMPUTE filter_$=(SCHOOLFEEDING = 1).
VARIABLE LABEL filter_$ 'SCHOOLFEEDING = 1 (FILTER)'.
VALUE LABELS filter_$ 0 'Not Selected' 1 'Selected'.
FORMAT filter_$ (f1.0).
FILTER BY filter_$.
EXECUTE.

 

T-TEST
  /TESTVAL=0
  /MISSING=ANALYSIS
  /VARIABLES=DiscrepancyHT1 Discrepancyteacher1
  /CRITERIA=CI(.95).

USE ALL.
COMPUTE filter_$=(SCHOOLFEEDING = 2).
VARIABLE LABEL filter_$ 'SCHOOLFEEDING = 2 (FILTER)'.
VALUE LABELS filter_$ 0 'Not Selected' 1 'Selected'.
FORMAT filter_$ (f1.0).
FILTER BY filter_$.
EXECUTE.

 

T-TEST
  /TESTVAL=0
  /MISSING=ANALYSIS
  /VARIABLES=DiscrepancyHT1 Discrepancyteacher1
  /CRITERIA=CI(.95).

[Art Kendall]

Please describe what your data looks like. 
What is a case (row)? 
How many teachers per head teacher?
Are the teachers and heads paired?
Do you have the 2 attendance variables (not just there discrepancy)?
Do you have multiple heads per school?
I am guessing that you have 3 "factors" (IVs):  factor A treatment vs control,  factor B which school,  factor C  head vs teacher.
Or are which head and which school the same thing? I.e., one head per school? In which situation, you have 2 factors.

Do you have an actual control group where there was random assignment to treatment vs control? or do you have a comparison group?  if you have a comparison group, how was it chosen?

Art Kendall
Social Research Consultants

On 10/27/2010 10:58 PM, Deepa Lakshmi Shanadi wrote:

Dear all,

 

1) I would like to do a ttest trying to see if there is a difference between headteachers and teachers  in their means on over/underreporting attendance.

 

2) Each HT and teacher belong to a intervention and control group.

 

3) HT and Teacher variables are continuous variables. There are negative numbers in the discrepancy colum (negative- underreporting, positive number- overreporting).

The syntax I have so far is:

 

T-TEST GROUPS=SCHOOLFEEDING(1 2)
  /MISSING=ANALYSIS
  /VARIABLES=Discrepancyteacher1 DiscrepancyHT1
  /CRITERIA=CI(.95).

However when I get my output, it looks like it is tell ing me if there is a difference between teachers in the intervention and control group and a difference between headteachers in the intervention and control group.

 

Am I using the wrong type of ttest?  Is this the ttest that I need- see below? And what do I need to be careful of with the negative numbers in the column? Thanks so much as usual?

 

USE ALL.
COMPUTE filter_$=(SCHOOLFEEDING = 1).
VARIABLE LABEL filter_$ 'SCHOOLFEEDING = 1 (FILTER)'.
VALUE LABELS filter_$ 0 'Not Selected' 1 'Selected'.
FORMAT filter_$ (f1.0).
FILTER BY filter_$.
EXECUTE.

 

T-TEST
  /TESTVAL=0
  /MISSING=ANALYSIS
  /VARIABLES=DiscrepancyHT1 Discrepancyteacher1
  /CRITERIA=CI(.95).

USE ALL.
COMPUTE filter_$=(SCHOOLFEEDING = 2).
VARIABLE LABEL filter_$ 'SCHOOLFEEDING = 2 (FILTER)'.
VALUE LABELS filter_$ 0 'Not Selected' 1 'Selected'.
FORMAT filter_$ (f1.0).
FILTER BY filter_$.
EXECUTE.

 

T-TEST
  /TESTVAL=0
  /MISSING=ANALYSIS
  /VARIABLES=DiscrepancyHT1 Discrepancyteacher1
  /CRITERIA=CI(.95).

===================== 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]

OK.  I am getting a clearer picture. You only have a comparison group but not a control group.
How do you know that the two groups of schools are similar?

Are the 7 grades each from different schools?   Do you have only 1 grade per school? How did you decide what grade to pick at each school?

You say that the head and the teacher are not paired, but you say you have one of each per school. Do you have any way of finding out what school they are from so that they can be paired?

Art Kendall
Social Research Consultants

On 10/28/2010 2:38 PM, Deepa Lakshmi Shanadi wrote:

Thank you for your response.

 

Responses in bold below.

---

From: Art Kendall [hidden email]
To: Deepa Lakshmi Shanadi [hidden email]
Cc: [hidden email]
Sent: Thu, October 28, 2010 8:04:46 AM
Subject: Re: [SPSSX-L] question on ttest

Please describe what your data looks like. 
What is a case (row)? 

 

all variables are continuous

HT count

Teacher count

actual count

discrepancy ht

discrpancy teacher

This exists for seven grades.  so there is each of this data for 7 grades.

 

This is the request:

"I think I want to be able to say that HT's reports are significantly over reported than are teachers, and to know if SF HTs overreport more than do NSF HTs.  Not by grade, but by mean of means of each group."

How many teachers per head teacher?

1 to 1, but lots of missing data

 

Are the teachers and heads paired?

 

No

Do you have the 2 attendance variables (not just there discrepancy)?

Yes, we have 1) actual count. Then we have 2) HT count and 3) Teacher count. Then we have 3) HT discrepancy.Discrepancy is the difference between HT count and actual count. 4) Teacher discrepancy- difference between teacher count and actual count.

 

Do you have multiple heads per school?

No.

 

I am guessing that you have 3 "factors" (IVs):  factor A treatment vs control,  factor B which school,  factor C  head vs teacher.
Or are which head and which school the same thing? I.e., one head per school? In which situation, you have 2 factors.

 

1 headteacher response per school

1 teacher response per school

Do you have an actual control group where there was random assignment to treatment vs control? or do you have a comparison group?  if you have a comparison group, how was it chosen?

 

there are 22 intervention schools

20 control schools.

It is not randomized if that is what you are asking.

Art Kendall
Social Research Consultants

On 10/27/2010 10:58 PM, Deepa Lakshmi Shanadi wrote:

Dear all,

 

1) I would like to do a ttest trying to see if there is a difference between headteachers and teachers  in their means on over/underreporting attendance.

 

2) Each HT and teacher belong to a intervention and control group.

 

3) HT and Teacher variables are continuous variables. There are negative numbers in the discrepancy colum (negative- underreporting, positive number- overreporting).

The syntax I have so far is:

 

T-TEST GROUPS=SCHOOLFEEDING(1 2)
  /MISSING=ANALYSIS
  /VARIABLES=Discrepancyteacher1 DiscrepancyHT1
  /CRITERIA=CI(.95).

However when I get my output, it looks like it is tell ing me if there is a difference between teachers in the intervention and control group and a difference between headteachers in the intervention and control group.

 

Am I using the wrong type of ttest?  Is this the ttest that I need- see below? And what do I need to be careful of with the negative numbers in the column? Thanks so much as usual?

 

USE ALL.
COMPUTE filter_$=(SCHOOLFEEDING = 1).
VARIABLE LABEL filter_$ 'SCHOOLFEEDING = 1 (FILTER)'.
VALUE LABELS filter_$ 0 'Not Selected' 1 'Selected'.
FORMAT filter_$ (f1.0).
FILTER BY filter_$.
EXECUTE.

 

T-TEST
  /TESTVAL=0
  /MISSING=ANALYSIS
  /VARIABLES=DiscrepancyHT1 Discrepancyteacher1
  /CRITERIA=CI(.95).

USE ALL.
COMPUTE filter_$=(SCHOOLFEEDING = 2).
VARIABLE LABEL filter_$ 'SCHOOLFEEDING = 2 (FILTER)'.
VALUE LABELS filter_$ 0 'Not Selected' 1 'Selected'.
FORMAT filter_$ (f1.0).
FILTER BY filter_$.
EXECUTE.

 

T-TEST
  /TESTVAL=0
  /MISSING=ANALYSIS
  /VARIABLES=DiscrepancyHT1 Discrepancyteacher1
  /CRITERIA=CI(.95).

===================== 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